Update app.py

app.py

@@ -5,7 +5,6 @@ import gradio as gr
 from PIL import Image
 from models_transformer_sd3 import SD3Transformer2DModel
 from pipeline_stable_diffusion_3_ipa import StableDiffusion3Pipeline
-import gc
 import os
 from huggingface_hub import login
 
@@ -14,6 +13,7 @@ login(TOKEN)
 
 model_path = 'stabilityai/stable-diffusion-3.5-large'
 ip_adapter_path = './ip-adapter.bin'
+##ipadapter_path = hf_hub_download(repo_id="InstantX/SD3.5-Large-IP-Adapter", filename="ip-adapter.bin")
 image_encoder_path = "google/siglip-so400m-patch14-384"
 
 transformer = SD3Transformer2DModel.from_pretrained(
@@ -22,7 +22,7 @@ transformer = SD3Transformer2DModel.from_pretrained(
 
 pipe = StableDiffusion3Pipeline.from_pretrained(
     model_path, transformer=transformer, torch_dtype=torch.bfloat16
-) ## For ZeroGPU no .to("cuda")
+).to("cuda")
 
 pipe.init_ipadapter(
     ip_adapter_path=ip_adapter_path, 
@@ -35,6 +35,13 @@ def randomize_seed_fn(seed: int, randomize_seed: bool) -> int:
         seed = random.randint(0, 2000)
     return seed
 
+def resize_img(image, max_size=1024):
+    width, height = image.size
+    scaling_factor = min(max_size / width, max_size / height)
+    new_width = int(width * scaling_factor)
+    new_height = int(height * scaling_factor)
+    return image.resize((new_width, new_height), Image.LANCZOS)
+
 @spaces.GPU() ## For ZeroGPU
 def create_image(image_pil,
                  prompt,
@@ -47,6 +54,9 @@ def create_image(image_pil,
                  target="Load only style blocks",
                  ):
     
+    if image_pil is None:
+        return None
+    
     if target !="Load original IP-Adapter":
         if target=="Load only style blocks":
             scale = {
@@ -63,12 +73,10 @@ def create_image(image_pil,
             }
         #pipe.set_ip_adapter_scale(scale) ## Waiting for SD3 Diffuser integration
     
-    if hasattr(image_pil, 'read'):  # If it's a file
-        style_image = Image.open(image_pil).convert('RGB')
-    else:  # If it's already a PIL Image
-        style_image = image_pil.convert('RGB')
+    if not isinstance(image_pil, Image.Image):  # If it's a file
+        image_pil = Image.fromarray(image_pil)
 
-    
+    image_pil = resize_img(image_pil)
     generator = torch.Generator().manual_seed(randomize_seed_fn(seed, True))
     
     image = pipe(
@@ -76,20 +84,10 @@ def create_image(image_pil,
             height=1024,
             prompt=prompt,
             negative_prompt="lowres, low quality, worst quality",
-            num_inference_steps=24, 
-            guidance_scale=guidance_scale,
             generator=generator,  ## For ZeroGPU no device="cpu"
-            clip_image=style_image,
-            ipadapter_scale=scale,
+            clip_image=image_pil,
+            ipadapter_scale=1,
         ).images[0]
-
-    
-    if torch.cuda.is_available():
-        torch.cuda.empty_cache()
-        gc.collect()
-
-    
-    print(image)
     
     return image