Update app.py

app.py

@@ -65,6 +65,24 @@ def gpt3_short(question,vqa_answer,caption):
     llm_ans=vqa_answer[ord(llm_ans)-ord("A")]
     answer=llm_ans
     
+    return answer
+def gpt3_long(question,vqa_answer,caption):
+    vqa_answer,vqa_score=vqa_answer
+    prompt="prompt: This is the caption of a picture: "+caption+". Question: "+question+" VQA model predicts:"+"A: "+vqa_answer[0]+"socre:"+str(vqa_score[0])+\
+           " B: "+vqa_answer[1]+" score:"+str(vqa_score[1])+" C: "+vqa_answer[2]+" score:"+str(vqa_score[2])+\
+            " D: "+vqa_answer[3]+'score:'+str(vqa_score[3])+\
+           "Tell me the right answer with a long sentence."
+    
+    # prompt=caption+"\n"+question+"\n"+vqa_answer+"\n Tell me the right answer."
+    response = openai.Completion.create(
+    engine="text-davinci-003",
+    prompt=prompt,
+    max_tokens=30,
+    n=1,
+    stop=None,
+    temperature=0.7,
+    )
+    answer = response.choices[0].text.strip()    
     return answer
 def gpt3(question,vqa_answer,caption):
     prompt=caption+"\n"+question+"\n"+vqa_answer+"\n Tell me the right answer."
@@ -94,8 +112,9 @@ def inference_chat(input_image,input_text):
     # out=processor.batch_decode(out, skip_special_tokens=True)
 
     out=vle(input_image,input_text)
-    vqa="\n".join(out[0])
-    gpt3_out=gpt3(input_text,vqa,cap)
+    # vqa="\n".join(out[0])
+    # gpt3_out=gpt3(input_text,vqa,cap)
+    gpt3_out=gpt3_long(input_text,out,cap)
     gpt3_out1=gpt3_short(input_text,out,cap)
     return out[0][0], gpt3_out,gpt3_out1
 title = """# VQA with VLE and LLM"""