Update app.py

app.py

@@ -24,11 +24,23 @@ llama = Llama(
 )
 
 
+
+
 # Function to generate responses
 def generate_response(message, history, system_prompt, temperature, max_new_tokens, top_k, repetition_penalty, top_p):
-    # chat_prompt = f"You are an Urdu Chatbot. Write an appropriate response for the given instruction: {message} Response:"
-    chat_prompt = f"{system_prompt}\n ### Instruction: {message}\n ### Response:"
-    response = llama(chat_prompt, temperature=temperature, max_tokens=max_new_tokens, top_k=top_k, repeat_penalty=repetition_penalty, top_p=top_p, stop=["Q:", "\n"], echo=False, stream=True)
+    
+    prompt_messages = system_prompt
+
+    # Add history and the current message
+    for user, bot in history:
+        prompt_messages = prompt_messages + f"\n### Instruction: {user}\n### Response: {bot}"
+        
+
+    prompt_messages = prompt_messages + f"\n### Instruction: {message}\n### Response: "
+    
+    print(prompt_messages)
+    
+    response = llama(prompt_messages, temperature=temperature, max_tokens=max_new_tokens, top_k=top_k, repeat_penalty=repetition_penalty, top_p=top_p, stop=["Q:", "\n"], echo=False, stream=True)
     
     text = ""
     for chunk in response:
@@ -37,29 +49,6 @@ def generate_response(message, history, system_prompt, temperature, max_new_toke
             text += content
             yield text
 
-# def generate_response(message, history, system_prompt, temperature, max_new_tokens, top_k, repetition_penalty, top_p):
-#     """Generates a streaming response from the Llama model."""
-#     messages = [
-#         {"role": "system", "content": "You are an Urdu Chatbot. Write an appropriate response for the given instruction."},
-#     ]
-
-#     # Add history and the current message
-#     #for user, bot in history:
-#         #messages.append({"role": "user", "content": user})
-#         #messages.append({"role": "assistant", "content": bot})
-
-#     messages.append({"role": "user", "content": message})
-
-#     response = llama.create_chat_completion(
-#         messages=messages,
-#         stream=True,
-#     )
-
-#     partial_message = ""
-#     for part in response:
-#         content = part["choices"][0]["delta"].get("content", "")
-#         partial_message += content
-#         yield partial_message
 
 
 # JavaScript function for `on_load`
@@ -98,7 +87,7 @@ with gr.Blocks(js=on_load, theme=gr.themes.Default()) as demo:
         additional_inputs=[
             gr.Textbox(value="You are an Urdu Chatbot. Write an appropriate response for the given instruction in Urdu.", label="System prompt", render=False),
             gr.Slider(0, 1, 0.8, label="Temperature", render=False),
-            gr.Slider(128, 4096, 512, label="Max new tokens", render=False),
+            gr.Slider(128, 4096, 1024, label="Max new tokens", render=False),
             gr.Slider(1, 80, 40, step=1, label="Top K sampling", render=False),
             gr.Slider(0, 2, 1.1, label="Repetition penalty", render=False),
             gr.Slider(0, 1, 0.95, label="Top P sampling", render=False),