updating v4

app.py

@@ -2,32 +2,29 @@ import streamlit as st
 import json
 import os
 import re
-import requests
 import uuid
 import time
 import torch
 from transformers import AutoModelForCausalLM, AutoTokenizer, pipeline
 
-# Load the Hugging Face model
-model_name = "ethicsadvisorproject/Llama-2-7b-ethical-chat-finetune"
-tokenizer = AutoTokenizer.from_pretrained(model_name, cache_dir="/tmp")
-model = AutoModelForCausalLM.from_pretrained(
-    model_name,
-    torch_dtype=torch.float16,
-    device_map="auto",
-    offload_folder="/tmp"
-)
-
-pipe = pipeline("text-generation", model=model, tokenizer=tokenizer, max_length=200)
-
+# Constants
+MODEL_NAME = "ethicsadvisorproject/Llama-2-7b-ethical-chat-finetune"
 DB_DIR = 'user_data'  # Directory to store individual user data
 os.makedirs(DB_DIR, exist_ok=True)  # Ensure the directory exists
 
-def stream_data(textto):
-    for word in textto.split(" "):
-        yield word + " "
-        time.sleep(0.02)
-        
+# Load the Hugging Face model
+@st.cache_resource
+def load_model():
+    tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME, cache_dir="/tmp")
+    model = AutoModelForCausalLM.from_pretrained(
+        MODEL_NAME,
+        torch_dtype=torch.float16,
+        device_map="auto",
+        offload_folder="/tmp"
+    )
+    return pipeline("text-generation", model=model, tokenizer=tokenizer, max_length=200)
+
+pipe = load_model()
 
 def get_user_id():
     """Generate or retrieve a unique ID for the user."""
@@ -36,7 +33,7 @@ def get_user_id():
     return st.session_state.user_id
 
 def get_user_file(user_id):
-    """Return the file path for a user's data file ok."""
+    """Return the file path for a user's data file."""
     return os.path.join(DB_DIR, f"{user_id}.json")
 
 def load_user_data(user_id):
@@ -54,47 +51,34 @@ def save_user_data(user_id, data):
         json.dump(data, file)
 
 def main():
-    endpoint_url = "https://zizytd-ethical-app-docker.hf.space/predict"  # Endpoint URL from .env
-
-    user_id = get_user_id()
-    user_data = load_user_data(user_id)
-
-#     st.set_page_config(page_title='Ethical GPT Assistant', layout='wide',
-#                 #    initial_sidebar_state=st.session_state.get('sidebar_state', 'collapsed'),
-# )
+    st.set_page_config(page_title='Ethical GPT Assistant', layout='wide')
 
-    st.image("./logo/images.jpeg", use_container_width=True )
-    
-    #st.snow()
+    st.image("./logo/images.jpeg", use_column_width=True)
 
     intro = """
     ## Welcome to EthicsAdvisor
-    
-    Ethical GPT is an AI-powered chatbot designed to interact with you in an ethical, safe, and responsible manner. Our goal is to ensure that all responses provided by the assistant are respectful and considerate of various societal and ethical standards.
 
+    Ethical GPT is an AI-powered chatbot designed to interact with you in an ethical, safe, and responsible manner. Our goal is to ensure that all responses provided by the assistant are respectful and considerate of various societal and ethical standards.
     Feel free to ask any questions, and rest assured that the assistant will provide helpful and appropriate responses.
     """
-    
     st.markdown(intro)
 
     # Sidebar options
-    models = ["llama-ethical"]
-    st.sidebar.selectbox("Select Model", models, index=0)
-
     st.sidebar.title("❄️EthicsAdvisor 📄")
-    st.sidebar.caption("Make AI to responde more ethical")
-    
-   
-    with st.sidebar.expander("See fine tuning info"):
-        st.caption("Original Data: [Data] (https://huggingface.co/datasets/MasahiroKaneko/eagle/)")
+    st.sidebar.caption("Make AI responses more ethical")
+
+    with st.sidebar.expander("See fine-tuning info"):
+        st.caption("Original Data: [Data](https://huggingface.co/datasets/MasahiroKaneko/eagle/)")
         st.caption("Modified Data: [Data](https://huggingface.co/datasets/ethicsadvisorproject/ethical_data_bigger/) 📝")
-        st.caption("Used Model and Notebook: [Original model](https://huggingface.co/ethicsadvisorproject/Llama-2-7b-ethical-chat-finetune/) 🎈, Notebook used for fine tuning [Notebook](https://colab.research.google.com/drive/1eAAjdwwD0i-i9-ehEJYUKXvZoYK0T3ue#scrollTo=ib_We3NLtj2E)")
-       
+        st.caption("Used Model and Notebook: [Original model](https://huggingface.co/ethicsadvisorproject/Llama-2-7b-ethical-chat-finetune/) 🎈, Notebook used for fine-tuning [Notebook](https://colab.research.google.com/drive/1eAAjdwwD0i-i9-ehEJYUKXvZoYK0T3ue#scrollTo=ib_We3NLtj2E)")
+
     with st.sidebar.expander("ℹ️ **Take survey**"):
-        st.markdown("""You are welcome to give us your input on this research [here](https://forms.office.com/r/H4ARtETV2q).""")
-    cols = st.columns(2)
+        st.markdown("You are welcome to give us your input on this research [here](https://forms.office.com/r/H4ARtETV2q).")
+
+    # Initialize chat history
+    user_id = get_user_id()
+    user_data = load_user_data(user_id)
 
-    # Load chat history into session state
     if "messages" not in st.session_state:
         st.session_state.messages = user_data["chat_history"]
 
@@ -103,44 +87,20 @@ def main():
         with st.chat_message(message["role"]):
             st.markdown(message["content"])
 
-    # # User input
-    # # if prompt := st.chat_input("What is up?"):
-    # #     st.session_state.messages.append({"role": "user", "content": prompt})
-    # #     with st.chat_message("user"):
-    # #         st.markdown(prompt)
-
-    # #     # Send request to the endpoint
-    # #     headers = {'ngrok-skip-browser-warning': 'true'}
-    # #     data = {'messages': st.session_state.messages[-1]['content']}
-
-    # #     try:
-    # #         response = requests.post(endpoint_url, json=data, headers=headers)
-    # #         response.raise_for_status()  # Raise exception for HTTP errors
-    # #         response_data = response.json()
-    # #         response_text = response_data.get('response_text', '')
+    # User input
+    prompt = st.chat_input("What is up?")
+    if prompt:
+        st.session_state.messages.append({"role": "user", "content": prompt})
+        with st.chat_message("user"):
+            st.markdown(prompt)
 
-    #         # Clean response text
-    #         message = re.sub(r'<s>\[INST\].*?\[/INST\]', '', response_text).strip()
+        # Generate response
+        response = pipe(f"<s>[INST] {prompt} [/INST]")
+        response_text = response[0]["generated_text"].replace("<s>[INST]", "").replace("[/INST]", "").strip()
 
-    #         with st.chat_message("assistant"):
-    #             st.markdown(message)
-
-    #         st.session_state.messages.append({"role": "assistant", "content": message})
-
-    #     except requests.exceptions.RequestException as e:
-    #         st.error(f"Error communicating with the endpoint: {e}")
-    #     except KeyError:
-    #         st.error(f"Unexpected response format. Missing 'response_text' key. Received: {response.text}")
-
-        if prompt := st.chat_input("What is up?"):
-            response = pipe(f"<s>[INST] {prompt} [/INST]")
-            response_text = response[0]["generated_text"].replace("<s>[INST]", "").replace("[/INST]", "").strip()
-
-        message = re.sub(r'<s>\[INST\].*?\[/INST\]', '', response_text).strip()
-    
         with st.chat_message("assistant"):
             st.markdown(response_text)
-    
+
         st.session_state.messages.append({"role": "assistant", "content": response_text})
 
         # Save updated chat history
@@ -152,7 +112,7 @@ def main():
         st.session_state.messages = []
         user_data["chat_history"] = []
         save_user_data(user_id, user_data)
-        st.rerun()
+        st.experimental_rerun()
 
 if __name__ == '__main__':
-    main()
+    main()