Update app.py

app.py

@@ -18,8 +18,8 @@ blip_model_base = BlipForQuestionAnswering.from_pretrained("Salesforce/blip-vqa-
 blip_processor_large = AutoProcessor.from_pretrained("Salesforce/blip-vqa-capfilt-large")
 blip_model_large = BlipForQuestionAnswering.from_pretrained("Salesforce/blip-vqa-capfilt-large")
 
-vilt_processor = AutoProcessor.from_pretrained("dandelin/vilt-b32-finetuned-vqa")
-vilt_model = ViltForQuestionAnswering.from_pretrained("dandelin/vilt-b32-finetuned-vqa")
+# vilt_processor = AutoProcessor.from_pretrained("dandelin/vilt-b32-finetuned-vqa")
+# vilt_model = ViltForQuestionAnswering.from_pretrained("dandelin/vilt-b32-finetuned-vqa")
 
 device = "cuda" if torch.cuda.is_available() else "cpu"
 
@@ -27,7 +27,7 @@ git_model_base.to(device)
 blip_model_base.to(device)
 git_model_large.to(device)
 blip_model_large.to(device)
-vilt_model.to(device)
+# vilt_model.to(device)
 
 def generate_answer_git(processor, model, image, question):
     # prepare image
@@ -41,7 +41,7 @@ def generate_answer_git(processor, model, image, question):
     generated_ids = model.generate(pixel_values=pixel_values, input_ids=input_ids, max_length=128)#50)
     generated_answer = processor.batch_decode(generated_ids, skip_special_tokens=True)
    
-    return generated_answer
+    return generated_answer.replace(question, '').replace(question.lower(), '').strip()
 
 
 def generate_answer_blip(processor, model, image, question):