update the design

app.py

@@ -20,13 +20,6 @@ headers = {
     "Authorization": f"Bearer {API_KEY}",
 }
 
-sys_msg = [
-    {
-        "content": "You are a helpful assistant. Be respectful, professional, and try to be helpful. If you don't know something, just say \"I don't know\"",
-        "role": "system",
-    }
-]
-
 
 def is_valid_json(data):
     try:
@@ -40,12 +33,19 @@ def is_valid_json(data):
 
 with gr.Blocks() as demo:
 
+    markup = gr.Markdown(
+        """
+                         # Mistral 7B Instruct v0.2
+                         This is a demo of the Mistral 7B Instruct quantized model in GGUF (Q2) hosted on K8s cluster.
+
+                         The original models can be found [MaziyarPanahi/Mistral-7B-Instruct-v0.2-GGUF](https://huggingface.co/MaziyarPanahi/Mistral-7B-Instruct-v0.2-GGUF)"""
+    )
     chatbot = gr.Chatbot()
-    msg = gr.Textbox()
+    msg = gr.Textbox(lines=1, label="User Message")
     clear = gr.Button("Clear")
     with gr.Row():
 
-        with gr.Column(scale=4):
+        with gr.Column(scale=2):
             # Define inputs for additional parameters
             system_prompt_input = gr.Textbox(
                 label="System Prompt",
@@ -58,27 +58,31 @@ with gr.Blocks() as demo:
             max_new_tokens_input = gr.Slider(
                 label="Max New Tokens", minimum=0, maximum=1024, value=256, step=1
             )
+
+        with gr.Column(scale=2):
             top_p_input = gr.Slider(
-                label="Top P", minimum=0.0, maximum=1.0, value=0.9, step=0.01
+                label="Top P", minimum=0.0, maximum=1.0, value=0.95, step=0.01
+            )
+            top_k_input = gr.Slider(
+                label="Top K", minimum=1, maximum=100, value=50, step=1
             )
             repetition_penalty_input = gr.Slider(
                 label="Repetition Penalty",
                 minimum=1.0,
                 maximum=2.0,
-                value=1.2,
+                value=1.1,
                 step=0.01,
             )
-        with gr.Column(scale=1):
-            markup = gr.Markdown("## Mistral 7B Instruct v0.2 GGUF")
 
     def update_globals(
-        system_prompt, temperature, max_new_tokens, top_p, repetition_penalty
+        system_prompt, temperature, max_new_tokens, top_p, top_k, repetition_penalty
     ):
-        global global_system_prompt, global_temperature, global_max_new_tokens, global_top_p, global_repetition_penalty
+        global global_system_prompt, global_temperature, global_max_new_tokens, global_top_p, global_repetition_penalty, global_top_k
         global_system_prompt = system_prompt
         global_temperature = temperature
         global_max_new_tokens = max_new_tokens
         global_top_p = top_p
+        global_top_k = top_k
         global_repetition_penalty = repetition_penalty
 
     def user(user_message, history):
@@ -87,13 +91,20 @@ with gr.Blocks() as demo:
         return "", history + [[user_message, None]]
 
     def bot(
-        history, system_prompt, temperature, max_new_tokens, top_p, repetition_penalty
+        history,
+        system_prompt,
+        temperature,
+        max_new_tokens,
+        top_p,
+        top_k,
+        repetition_penalty,
     ):
         print(f"History in bot: {history}")
         print(f"System Prompt: {system_prompt}")
         print(f"Temperature: {temperature}")
         print(f"Max New Tokens: {max_new_tokens}")
         print(f"Top P: {top_p}")
+        print(f"Top K: {top_k}")
         print(f"Repetition Penalty: {repetition_penalty}")
 
         # print(f"History in bot: {history}")
@@ -103,7 +114,14 @@ with gr.Blocks() as demo:
         # let's extract the user's question which should be the last touple first element
         # user_question = history[-1][0]
         history[-1][1] = ""
-
+        sys_msg = [
+            {
+                "content": (
+                    system_prompt if system_prompt else "You are a helpful assistant."
+                ),
+                "role": "system",
+            }
+        ]
         history_messages = sys_msg + history_messages
         print(history_messages)
 
@@ -111,8 +129,8 @@ with gr.Blocks() as demo:
             "messages": history_messages,
             "stream": True,
             "temprature": temperature,
-            "top_k": 50,
-            "top_p": 0.95,
+            "top_k": top_k,
+            "top_p": top_p,
             "seed": 42,
             "repeat_penalty": repetition_penalty,
             "chat_format": "mistral-instruct",
@@ -167,6 +185,7 @@ with gr.Blocks() as demo:
             temperature_input,
             max_new_tokens_input,
             top_p_input,
+            top_k_input,
             repetition_penalty_input,
         ],
         outputs=chatbot,