Update app.py

app.py

@@ -29,16 +29,19 @@ def process(model_A, model_B):
     prompt = sample["prompt"]
     
     df = pd.DataFrame.from_records(sample["filtered_outputs"])
-    response_A = df[df['model']==model_A]["output"]
+    response_A_df = df[df['model']==model_A]["output"]
+    response_B_df= df[df['model']==model_B]["output"]
+
+    response_A = response_A_df.values
+    response_B = response_B_df.values
     print(response_A)
-    response_B = df[df['model']==model_B]["output"]
     
         
     input_text = "POST: "+ prompt+ "\n\n RESPONSE A: "+response_A+"\n\n RESPONSE B: "+response_B+"\n\n Which response is better? RESPONSE"
     x = tokenizer([input_text], return_tensors='pt').input_ids.to(device)
     y = model.generate(x, max_new_tokens=1)
     prefered = tokenizer.batch_decode(y, skip_special_tokens=True)[0] 
-    return pd.concat([response_A,response_B]), prefered
+    return pd.concat([response_A_df,response_B_df]), prefered
 
 title = "Compare Instruction Models to see which one is more helpful"
 interface = gr.Interface(fn=process,