Update app.py

app.py

@@ -64,8 +64,6 @@ def generate_response(
     max_new_tokens: int,
     top_p: float,
     top_k: int,
-    repetition_penalty: float,
-    no_repeat_ngram_size: int,
 ):
     system_prompt = """You are a Kamala Harris chatbot. You only answer like Harris in style and tone. In every response:
 1. Maintain a composed and professional demeanor.
@@ -96,8 +94,6 @@ Crucially, Keep responses concise and impactful."""
             top_p=top_p,
             top_k=top_k,
             temperature=temperature,
-            repetition_penalty=repetition_penalty,
-            no_repeat_ngram_size=no_repeat_ngram_size,
             pad_token_id=tokenizer.pad_token_id,
             eos_token_id=tokenizer.eos_token_id,
         )
@@ -109,9 +105,9 @@ def add_text(history, text):
     history = history + [(text, None)]
     return history, ""
 
-def bot(history, temperature, max_new_tokens, top_p, top_k, repetition_penalty, no_repeat_ngram_size):
+def bot(history, temperature, max_new_tokens, top_p, top_k):
     user_message = history[-1][0]
-    bot_response = generate_response(user_message, history[:-1], temperature, max_new_tokens, top_p, top_k, repetition_penalty, no_repeat_ngram_size)
+    bot_response = generate_response(user_message, history[:-1], temperature, max_new_tokens, top_p, top_k)
     history[-1][1] = bot_response
     return history
 
@@ -136,11 +132,9 @@ with gr.Blocks(css=CSS, theme=gr.themes.Default()) as demo:
 
     with gr.Accordion("Advanced Settings", open=False):
         temperature = gr.Slider(minimum=0.1, maximum=1.5, value=0.8, step=0.1, label="Temperature")
-        max_new_tokens = gr.Slider(minimum=50, maximum=500, value=256, step=1, label="Max New Tokens")
-        top_p = gr.Slider(minimum=0.1, maximum=1.0, value=0.9, step=0.1, label="Top-p")
+        max_new_tokens = gr.Slider(minimum=50, maximum=1024, value=1024, step=1, label="Max New Tokens")
+        top_p = gr.Slider(minimum=0.1, maximum=1.5, value=1.0, step=0.1, label="Top-p")
         top_k = gr.Slider(minimum=1, maximum=100, value=20, step=1, label="Top-k")
-        repetition_penalty = gr.Slider(minimum=1.0, maximum=2.0, value=1.2, step=0.1, label="Repetition Penalty")
-        no_repeat_ngram_size = gr.Slider(minimum=1, maximum=10, value=3, step=1, label="No Repeat N-gram Size")
 
     gr.Examples(
         examples=[
@@ -152,11 +146,11 @@ with gr.Blocks(css=CSS, theme=gr.themes.Default()) as demo:
     )
 
     submit.click(add_text, [chatbot, msg], [chatbot, msg], queue=False).then(
-        bot, [chatbot, temperature, max_new_tokens, top_p, top_k, repetition_penalty, no_repeat_ngram_size], chatbot
+        bot, [chatbot, temperature, max_new_tokens, top_p, top_k], chatbot
     )
     clear.click(lambda: [], outputs=[chatbot], queue=False)
     msg.submit(add_text, [chatbot, msg], [chatbot, msg], queue=False).then(
-        bot, [chatbot, temperature, max_new_tokens, top_p, top_k, repetition_penalty, no_repeat_ngram_size], chatbot
+        bot, [chatbot, temperature, max_new_tokens, top_p, top_k], chatbot
     )
 
 if __name__ == "__main__":