Update app.py

app.py

@@ -36,10 +36,16 @@ with st.container():
 
     #Check if chat history exists in this session
     if 'chat_history' not in st.session_state:
-        st.session_state.chat_history = [ ] #Initialize chat history
+        st.session_state.chat_history = [
+            {
+                "role": "system",
+                "content": "You are a friendly chatbot who always helps the user book a hotel room based on his/her needs."
+                + "Based on the current social norms you wait for the user's response to your proposals.",
+            },
+            {"role": "assistant", "content": "Hello, how can I help you today?"},
+        ] #Initialize chat history
 
     if 'model' not in st.session_state:
-        st.write("Model added in state.")
         st.session_state.model = model
 
     #renders chat history
@@ -47,7 +53,8 @@ with st.container():
         with st.chat_message(message["role"]):
             st.write(message["content"])
             
-
+    tokenized_chat = tokenizer.apply_chat_template(messages, tokenize=True, add_generation_prompt=True, return_tensors="pt")
+    st.write(tokenizer.decode(tokenized_chat[0]))
     #Set up input text field
     input_text = st.chat_input(placeholder="Here you can chat with our hotel booking model.")
 
@@ -58,8 +65,8 @@ with st.container():
 
         #chat_response = demo_chat.demo_chain(input_text=input_text, memory=st.session_state.memory, model= chat_model)
         #first_answer = chat_response.split("Human")[0] #Because of Predict it prints the whole conversation.Here we seperate the first answer only.
-        output = model_pipeline(input_text, max_new_tokens=20)
-        first_answer = output[0]["generated_text"]
+        outputs = model.generate(tokenized_chat, max_new_tokens=128) 
+        first_answer = tokenizer.decode(outputs[0])
         
         with st.chat_message("assistant"):
             st.write(first_answer)