Update app.py

app.py

@@ -152,7 +152,7 @@ def generate(
 
     # with torch.cuda.amp.autocast(dtype=torch.bfloat16):
     output = model.generate(
-        vision_x=vision_x.to("cuda", dtype=torch.bfloat16)
+        vision_x=vision_x.to("cuda", dtype=torch.bfloat16),
         lang_x=input_ids.to("cuda"),
         attention_mask=attention_mask.to("cuda"),
         max_new_tokens=30,