Update app.py

app.py

@@ -29,23 +29,24 @@ def process(model_A, model_B):
     prompt = sample["prompt"]
     
     df = pd.DataFrame.from_records(sample["filtered_outputs"])
-    response_A = df[df['model']==model_A]
-    response_B = df[df['model']==model_B]
+    response_A = df[df['model']==model_A]["output"]
+    response_B = df[df['model']==model_B]["output"]
     
         
-    input_text = "POST: "+ prompt+ "\n\n RESPONSE A: Lime juice, and zest, then freeze in small quantities.\n\n RESPONSE B: Lime marmalade lol\n\n Which response is better? RESPONSE"
+    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 prefered, response_A
+    return pd.concat([response_A,response_B], prefered
 
 title = "Compare Instruction Models to see which one is more helpful"
 interface = gr.Interface(fn=process, 
                      inputs=[gr.Dropdown(choices=model_list, value=model_list[0], label='Model A'),
                             gr.Dropdown(choices=model_list, value=model_list[1], label='Model B')],
                      outputs=[
+                              
+                              gr.DataFrame(label = "Model Responses"),
                               gr.Textbox(label = "Preferred Option"),
-                              gr.DataFrame(label = "Model Responses")
                                  
                               ],
                      title=title,