testing global preprocessor

app.py

@@ -110,9 +110,11 @@ pipe.to("cuda")
 
 print("loading preprocessor")
 preprocessor = Preprocessor()
-# preprocessor.load("NormalBae")
+preprocessor.load("NormalBae")
 
 print("---------------Loaded controlnet pipeline---------------") 
+print(f"CUDA memory allocated: {torch.cuda.max_memory_allocated(device='cuda') / 1e9:.2f} GB")
+
 
 # @spaces.GPU(duration=12)
 # def init(pipe):
@@ -401,6 +403,8 @@ def process_image(
     torch.cuda.synchronize()
     preprocess_start = time.time()
     print("processing image")
+
+    global preprocessor
     preprocessor.load("NormalBae")
 
     global compiled