Update app.py

app.py

@@ -17,14 +17,15 @@ def load_model():
         )
         if torch.cuda.is_available():
             pipe.to("cuda")
-            # Enable memory optimizations without CPU offload
+            # Enable memory optimizations
             pipe.enable_attention_slicing()
+            pipe.enable_model_cpu_offload()
     return pipe
 
 def generate_image(prompt, negative_prompt, steps=30, guidance_scale=7.5):
     pipe = load_model()
     
-    # Generate image with memory optimizations
+    # Generate image with reduced resolution for stability
     image = pipe(
         prompt=prompt,
         negative_prompt=negative_prompt,