add params to original

app.py

@@ -11,8 +11,22 @@ tokenizer = GPT2Tokenizer.from_pretrained('microsoft/DialoGPT-medium')
 original_model = GPT2LMHeadModel.from_pretrained('microsoft/DialoGPT-medium')
 fine_tuned_model = GPT2LMHeadModel.from_pretrained('zmbfeng/FineTune-1')
 
-def create_response_original(input_str, num_beams, input_num):
-  print ("input_num="+str(num))
+def create_response_original(input_str, 
+                             num_beams, 
+                             num_return_sequences,
+                             temperature,
+                             repetition_penalty,
+                             top_p,
+                             top_k,
+                             do_sample):
+  print("num_beams=" + num_beams)
+  print("num_return_sequences" + num_return_sequences)
+  print("top_p" + top_p)
+  print("top_k" + top_k)
+  print("repetition_penalty" + repetition_penalty)
+  print("temperature" + temperature)
+  print("do_sample" + do_sample)
+
   #output_raw= generator(input_str)
   """print (output_raw)"""
 
@@ -37,7 +51,32 @@ interface1 = gr.Interface(fn=create_response_original, inputs=[
     gr.Textbox(label="input text here", lines=3),
     gr.Number(label="num_beams (integer) explores the specified number of possible outputs and selects the most " +
                 "likely ones (specified in num_beams)", value=7),
-    gr.inputs.Number(label="Enter a whole number")], outputs="text", title="Original")
+    gr.Number(label="num_return_sequences (integer)  the number of outputs selected from num_beams possible output",
+                  value=5),
+        gr.Number(
+            label="temperature (decimal) controls the creativity or randomness of the output. A higher temperature" +
+                  " (e.g., 0.9) results in more diverse and creative output, while a lower temperature (e.g., 0.2)" +
+                  " makes the output more deterministic and focused",
+            value=0.2),
+        gr.Number(label="repetition_penalty (decimal) penalizes words that have already appeared in the output, " +
+                        "making them less likely to be generated again. A higher repetition_penalty (e.g., 1.5) results" +
+                        "in more varied and non-repetitive output.",
+                  value=1.5),
+        gr.Number(label="top_p (decimal) the model will only consider the words that have a high enough probability" +
+                        " to reach a certain threshold",
+                  value=0.9),
+        gr.Number(label="top_k (integer) The number of highest probability vocabulary word will be considered" +
+                        "This means that only the tokens with the highest probabilities are considered for sampling" +
+                        "This reduces the diversity of the generated sequences, "+
+                        "but also makes them more likely to be coherent and fluent.",
+                  value=50),
+        gr.Checkbox(label="do_sample. If is set to False, num_return_sequences must be 1 because the generate function will use greedy decoding, " +
+                          "which means that it will select the word with the highest probability at each step. " +
+                          "This results in a deterministic and fluent output, but it might also lack diversity and creativity" +
+                          "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", title="Original")
 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: