try something

app.py

@@ -116,8 +116,8 @@ def infer(prompt, seed=42, randomize_seed=False, width=1024, height=1024, num_in
         
         generator = torch.Generator("cuda").manual_seed(seed)  # Explicitly use CUDA generator
         
-        with torch.autocast('cuda'):
-            output = pipe(
+       # with torch.autocast('cuda'):
+        output = pipe(
                 prompt=prompt, 
                 width=width,
                 height=height,
@@ -128,13 +128,13 @@ def infer(prompt, seed=42, randomize_seed=False, width=1024, height=1024, num_in
             )
             
             # Ensure the image is properly normalized and converted
-            image = output.images[0]
-            if isinstance(image, torch.Tensor):
-                image = (image.clamp(-1, 1) + 1) / 2
-                image = (image * 255).round().clamp(0, 255).to(torch.uint8).cpu().numpy()
-                image = Image.fromarray(image)
+        image = output.images[0]
+        #    if isinstance(image, torch.Tensor):
+        #        image = (image.clamp(-1, 1) + 1) / 2
+        #        image = (image * 255).round().clamp(0, 255).to(torch.uint8).cpu().numpy()
+        #        image = Image.fromarray(image)
 
-        torch.cuda.empty_cache()  
+        #torch.cuda.empty_cache()  
         return image, seed
     except Exception as e:
         print(f"Error in infer: {str(e)}")