Update app.py

app.py

@@ -36,9 +36,9 @@ import os
 openai.api_key= os.getenv('openai_appkey') 
 def gpt3_short(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])+\
+    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])+\
            ". Choose A if it is not in conflict with the description of the picture and A's score is bigger than 0.8; otherwise choose the B, C or D based on the description. Answer with A or B or C or D."
     
     # prompt=caption+"\n"+question+"\n"+vqa_answer+"\n Tell me the right answer."
@@ -72,14 +72,14 @@ def gpt3_long(question,vqa_answer,caption):
     #        " 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="This is the caption of a picture: "+caption+". Question: "+question+" VQA model predicts:"+" "+vqa_answer[0]+", socre:"+str(vqa_score[0])+\
+           ";   "+vqa_answer[1]+", score:"+str(vqa_score[1])+";  "+vqa_answer[2]+", score:"+str(vqa_score[2])+\
+            ";  "+vqa_answer[3]+', score:'+str(vqa_score[3])+\
+           ". Question: "+question+" Tell me the right answer with a sentence."
     # prompt="prompt: This is the caption of a picture: "+caption+". Question: "+question+" VQA model predicts:"+" "+vqa_answer[0]+" socre:"+str(vqa_score[0])+\
     #        "   "+vqa_answer[1]+" score:"+str(vqa_score[1])+"  "+vqa_answer[2]+" score:"+str(vqa_score[2])+\
     #         "  "+vqa_answer[3]+'score:'+str(vqa_score[3])+\
-    #        "Tell me the right answer with a sentence."
-    prompt="prompt: This is the caption of a picture: "+caption+". Question: "+question+" VQA model predicts:"+" "+vqa_answer[0]+" socre:"+str(vqa_score[0])+\
-           "   "+vqa_answer[1]+" score:"+str(vqa_score[1])+"  "+vqa_answer[2]+" score:"+str(vqa_score[2])+\
-            "  "+vqa_answer[3]+'score:'+str(vqa_score[3])+\
-           "Tell me the right answer with a long sentence."
+    #        "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",