Spaces:
Running
Running
| import os | |
| import openai | |
| import streamlit as st | |
| from langchain.agents import AgentType, initialize_agent, load_tools | |
| from langchain.llms import OpenAI as l_OpenAI | |
| from transformers import pipeline | |
| from transformers import AutoTokenizer, AutoModelForCausalLM | |
| from helpers.foundation_models import * | |
| import requests | |
| # API Keys | |
| OPENAI_API_KEY = os.environ["OPENAI_API_KEY"] | |
| SERPAPI_API_KEY = os.environ["SERPAPI_API_KEY"] | |
| openai_client = openai.OpenAI(api_key=OPENAI_API_KEY) | |
| # Initialize chat history | |
| if "messages" not in st.session_state: | |
| st.session_state.messages = [] | |
| # Display chat messages from history on app rerun | |
| for message in st.session_state.messages: | |
| with st.chat_message(message["role"]): | |
| st.markdown(message["content"]) | |
| # Sidebar | |
| st.sidebar.markdown( | |
| r""" | |
| # 🌟 Streamlit + Hugging Face Demo 🤖 | |
| ## Introduction 📖 | |
| This demo showcases how to interact with Large Language Models (LLMs) on Hugging Face using Streamlit. | |
| """ | |
| ) | |
| option = st.sidebar.selectbox( | |
| "Which task do you want to do?", | |
| ("Sentiment Analysis", "Medical Summarization", "Llama2 on YSA", "Llama2 on BRK Letters", "ChatGPT", "ChatGPT (with Google)"), | |
| ) | |
| clear_button = st.sidebar.button("Clear Conversation", key="clear") | |
| st.sidebar.write("---") | |
| st.sidebar.markdown("Yiqiao Yin: [Site](https://www.y-yin.io/) | [LinkedIn](https://www.linkedin.com/in/yiqiaoyin/)") | |
| st.sidebar.markdown( | |
| r""" | |
| To fine-tune LLM such as Llama2 on custom data, please use the following tutorials as resources. For the options above such as `Llama2 on YSA` and `Llama2 on BRK Letters`, they are developed based on the content of the following videos. | |
| ## Video Series Overview | |
| ### Video 1: Process Your Own PDF Doc into LLM Finetune-Ready Format | |
| Learn how to transform PDF documents into AI model fine-tuning ready formats. This video will take you through the steps to make your PDF data AI-ready. | |
| - [Watch Video](https://youtu.be/hr2kSC1evQM) | |
| - [Tutorial Notebook](https://github.com/yiqiao-yin/WYNAssociates/blob/main/docs/ref-deeplearning/ex24f%20-%20process%20custom%20data%20from%20pdf%20and%20push%20to%20huggingface%20to%20prep%20for%20fine%20tune%20task%20of%20llama%202%20using%20lora.ipynb) | |
| ### Video 2: Fine-tune Llama2-7b LLM Using Custom Data | |
| Dive into customizing the Llama-2 model with your unique dataset. This installment turns your data into a bespoke AI model. | |
| - [Watch Video](https://youtu.be/tDkY2gpvylE) | |
| - [Guide to Fine-Tuning](https://github.com/yiqiao-yin/WYNAssociates/blob/main/docs/ref-deeplearning/ex24f%20-%20fine%20tune%20Llama%202%20using%20ysa%20data%20in%20colab.ipynb) | |
| ### Video 3: Deploy Inference Endpoint on HuggingFace! | |
| Discover how to make your AI model accessible to the world by deploying it on HuggingFace. This video turns your project into a global phenomenon. | |
| - [Watch Video](https://youtu.be/382yy-mCeCA) | |
| - [Deployment Guide](https://github.com/yiqiao-yin/WYNAssociates/blob/main/docs/ref-deeplearning/ex24f%20-%20inference%20endpoint%20interaction%20from%20huggingface.ipynb) | |
| - [HuggingFace Space](https://huggingface.co/spaces/eagle0504/streamlit-demo) | |
| """ | |
| ) | |
| # Reset everything | |
| if clear_button: | |
| st.session_state.messages = [] | |
| # React to user input | |
| if prompt := st.chat_input("What is up?"): | |
| # Display user message in chat message container | |
| st.chat_message("user").markdown(prompt) | |
| # Add user message to chat history | |
| st.session_state.messages.append({"role": "user", "content": prompt}) | |
| # Execute options | |
| with st.spinner("Wait for it..."): | |
| if option == "Sentiment Analysis": | |
| pipe_sentiment_analysis = pipeline("sentiment-analysis") | |
| if prompt: | |
| out = pipe_sentiment_analysis(prompt) | |
| final_response = f""" | |
| Prompt: {prompt} | |
| Sentiment: {out[0]["label"]} | |
| Score: {out[0]["score"]} | |
| """ | |
| elif option == "Medical Summarization": | |
| pipe_summarization = pipeline( | |
| "summarization", model="Falconsai/medical_summarization" | |
| ) | |
| if prompt: | |
| out = pipe_summarization(prompt) | |
| final_response = out[0]["summary_text"] | |
| elif option == "Llama2 on YSA": | |
| if prompt: | |
| try: | |
| out = llama2_7b_ysa(prompt) | |
| engineered_prompt = f""" | |
| The user asked the question: {prompt} | |
| We have found relevant content: {out} | |
| Answer the user question based on the above content in paragraphs. | |
| """ | |
| final_response = call_chatgpt(query=engineered_prompt) | |
| except: | |
| final_response = "Sorry, the inference endpoint is temporarily down. 😔" | |
| elif option == "Llama2 on BRK Letters": | |
| if prompt: | |
| try: | |
| out = llama2_7b_brk_letters(prompt) | |
| engineered_prompt = f""" | |
| The user asked the question: {prompt} | |
| We have found relevant content: {out} | |
| Answer the user question based on the above content in paragraphs. | |
| """ | |
| final_response = call_chatgpt(query=engineered_prompt) | |
| except: | |
| final_response = "Sorry, the inference endpoint is temporarily down. 😔" | |
| elif option == "ChatGPT": | |
| if prompt: | |
| out = call_chatgpt(query=prompt) | |
| final_response = out | |
| elif option == "ChatGPT (with Google)": | |
| if prompt: | |
| ans_langchain = call_langchain(prompt) | |
| prompt = f""" | |
| Based on the internet search results: {ans_langchain}; | |
| Answer the user question: {prompt} | |
| """ | |
| out = call_chatgpt(query=prompt) | |
| final_response = out | |
| else: | |
| final_response = "" | |
| response = f"{final_response}" | |
| # Display assistant response in chat message container | |
| with st.chat_message("assistant"): | |
| st.markdown(response) | |
| # Add assistant response to chat history | |
| st.session_state.messages.append({"role": "assistant", "content": response}) | |