Update app.py

app.py

@@ -2,23 +2,13 @@ import os
 import streamlit as st
 import chatbot as demo_chat
 from transformers import AutoModelForCausalLM, AutoTokenizer
-from langchain.schema import (
-    HumanMessage,
-    SystemMessage,
-)
-from langchain_community.chat_models.huggingface import ChatHuggingFace
-from transformers import pipeline
 
 st.title("Hi, I am Chatbot Philio :mermaid:")
 st.write("I am your hotel booking assistant for today.")
 
 tokenizer, model = demo_chat.load_model()
 
-model_identifier = "KvrParaskevi/Hotel-Assistant-Attempt4-Llama-2-7b"
-task = "text-generation"  # Change this to your model's task
-
-# Load the model using the pipeline
-model_pipeline = pipeline(task, model=model,tokenizer=tokenizer)
+model_identifier = "KvrParaskevi/Hotel-Assistant-Attempt5-Llama-2-7b"
 
 scrollable_div_style = """
 <style>
@@ -39,6 +29,13 @@ def render_chat_history(chat_history):
             with st.chat_message(message["role"]):
                 st.markdown(message["content"])
 
+def generate_response(chat_history):
+    tokenized_chat = tokenizer.apply_chat_template(chat_history, tokenize=True, add_generation_prompt=True, return_tensors="pt")
+    outputs = model.generate(tokenized_chat, do_sample =True, max_new_tokens=50, temperature = 0.5, top_p = 0.8)
+    answer = tokenizer.decode(outputs[0][tokenized_chat.shape[1]:],skip_special_tokens=True)
+    final_answer = answer.split("<")[0]
+    return final_answer
+
 #Application 
 #Langchain memory in session cache 
 if 'memory' not in st.session_state:
@@ -58,12 +55,10 @@ if 'chat_history' not in st.session_state:
 if 'model' not in st.session_state:
     st.session_state.model = model
     
-st.markdown('<div class="scrollable-div">', unsafe_allow_html=True)
+st.markdown('<div class="scrollable-div">', unsafe_allow_html=True) #add css style to container
 render_chat_history(st.session_state.chat_history)
 
-#Set up input text field
-#input_text = st.chat_input(placeholder="Here you can chat with our hotel booking model.")
-
+#Input field for chat interface
 if input_text := st.chat_input(placeholder="Here you can chat with our hotel booking model."):
     
     with st.chat_message("user"):
@@ -71,12 +66,7 @@ if input_text := st.chat_input(placeholder="Here you can chat with our hotel boo
     st.session_state.chat_history.append({"role" : "user", "content" : input_text}) #append message to chat history
 
     with st.spinner("Generating response..."):
-        #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.
-        tokenized_chat = tokenizer.apply_chat_template(st.session_state.chat_history, tokenize=True, add_generation_prompt=True, return_tensors="pt")
-        #st.write(tokenizer.decode(tokenized_chat[0]))
-        outputs = model.generate(tokenized_chat, max_new_tokens=128) 
-        first_answer = tokenizer.decode(outputs[0][tokenized_chat.shape[1]:],skip_special_tokens=True)
+        first_answer = generate_response(st.session_state.chat_history)
         
         with st.chat_message("assistant"):
             st.markdown(first_answer)