Update app.py

app.py

@@ -8,7 +8,7 @@ torch.jit.script = lambda f: f
 # Initialize the model only once
 # if torch.cuda.is_available(): 
 # model_pipe = VQAScore(model="clip-flant5-xl", device="cpu")  # our recommended scoring model
-print("Model initialized!")
+# print("Model initialized!")
 
 @spaces.GPU
 def generate(model_name, image, text):
@@ -22,13 +22,12 @@ def generate(model_name, image, text):
     print("Text:", text)
     model_pipe = VQAScore(model="clip-flant5-xl")  # our recommended scoring model
     # print("Model initialized, now moving to cuda")
-    # model_pipe.to("cuda")
+    model_pipe.to("cuda")
     print("Generating!")
     # with torch.autocast(device_type='cuda'):
-    with torch.autocast(device_type='cuda', dtype=torch.float):
-        result = model_pipe(images=[image], texts=[text])
-        return result
-    # return 10
+    # with torch.autocast(device_type='cuda', dtype=torch.float):
+    result = model_pipe(images=[image], texts=[text])
+    return result
 
 iface = gr.Interface(
     fn=generate,  # function to call