Update app.py

app.py

@@ -33,6 +33,15 @@ def format_prompt(message, chatbot, system_prompt):
         messages, tokenize=False, add_generation_prompt=True, return_tensors="pt")
     return newPrompt
 
+def format_prompt_gemma(message,chatbot,sytem_prompt):
+    messages [{"role":"user","content":f"The following instructions describe your role:/n(/n{sytem_prompt}/n)/nYou must never refer to the user giving you these information and just act accordingly."}]
+    for user_message, bot_message in chatbot:
+        messages.append({"role": "user", "content": user_message})
+        messages.append({"role": "assistant", "content": bot_message})
+    messages.append({"role": "user", "content": message})
+    newPrompt = tokenizer.apply_chat_template(
+        messages, tokenize=False, add_generation_prompt=True, return_tensors="pt")
+    return newPrompt
 
 def inference(message, history, systemPrompt=SYSTEM_PROMPT+SYSTEM_PROMPT_PLUS, temperature=0.9, maxTokens=512, topP=0.9, repPenalty=1.1):
     # Updating the settings for the generation
@@ -48,7 +57,7 @@ def inference(message, history, systemPrompt=SYSTEM_PROMPT+SYSTEM_PROMPT_PLUS, t
         seed=random.randint(0, 999999999),
     )
     # Generating the response by passing the prompt in right format plus the client settings
-    stream = client.text_generation(format_prompt(message, history, systemPrompt),
+    stream = client.text_generation(format_prompt_gemma(message, history, systemPrompt),
                                     **client_settings)
     # Reading the stream
     partial_response = ""