first try using param and multiple output

app.py

@@ -34,8 +34,25 @@ def create_response_original(input_str,
   #output_str = output_str.replace("\n", "")
   #output_str = output_str.replace(input_str, "")
   #output_str = tokenizer.decode(model.generate(**tokenizer("What are John West's hobbies?"+tokenizer.eos_token,return_tensors="pt",max_length=200))[0])
-  output_str = tokenizer.decode(original_model.generate(**tokenizer(input_str+tokenizer.eos_token,return_tensors="pt",max_length=200))[0])
-  return (output_str)
+  # output_str = tokenizer.decode(original_model.generate(**tokenizer(input_str+tokenizer.eos_token,return_tensors="pt",max_length=200),
+  #                                                      num_beams=num_beams,
+  #                                                      num_return_sequences=num_return_sequences)[0])
+
+  input_ids = tokenizer.encode(input_str + tokenizer.eos_token, return_tensors="pt")
+  output_ids = original_model.generate(input_ids,
+                                           do_sample=do_sample, 
+                                           max_length=100, 
+                                           temperature=temperature,  
+                                           top_p=top_p,
+                                           top_k=top_k,
+                                           repetition_penalty=repetition_penalty,
+                                           num_return_sequences=num_return_sequences, 
+                                           num_beams=num_beams)                                 
+  outputs = []]                                 
+  for output_id in output_ids:
+    output = tokenizer.decode(output_id, skip_special_tokens=True)
+    outputs.append(output)
+  return outputs
 
 def create_response_fine_tuned(input_str):
   #output_raw= generator(input_str)
@@ -85,7 +102,7 @@ interface1 = gr.Interface(fn=create_response_original,
                           "If is set to True, the generate function will use stochastic sampling, which means that it will randomly" +
                           " select a word from the probability distribution at each step. This results in a more diverse and creative" +
                           " output, but it might also introduce errors and inconsistencies ", value=True)
-    ], outputs="text")
+    ], outputs="list")
 interface2 = gr.Interface(fn=create_response_fine_tuned, inputs="text", outputs="text", title="Fine Tuned")
 demo = gr.TabbedInterface([interface1, interface2], ["Original", "Fine Tuned"])
 # with gr.Blocks() as demo: