Τraining modules for SMEs - Module 3 "Introduction to Large Language Models"

SmartAttica EDIH (European Digital Innovation Hub) announces the 3rd Module of Τraining modules for SMEs with the subject "Developing your Custom Chatbot with Ease", that will take place online on March 14th, 2025.   

Date: March 14th, 2025, at 11:00 EET  

Location: Online via Zoom

Presentation Languages: Greek

Instructor: Dr. Nikolaos Bakas (GRNET)

Description: This is a hands-on introduction to building advanced chatbots using HuggingFace and Retrieval-Augmented Generation (RAG). It covers the basics of pre-trained model selection, integration with Streamlit for user-friendly interfaces, and using RAG to enhance chatbot capabilities with external information sources. Participants will gain practical skills in setting up and customizing chatbots that are both responsive and contextually aware.

Target Audience:

This seminar module is designed for small and medium-sized enterprise (SME) developers, technical leads, and data scientists who are interested in incorporating natural language processing (NLP) capabilities into their projects. It is ideal for those looking to quickly prototype and deploy intelligent conversational agents using cutting-edge AI technologies.

Learning Objectives:

By the end of this module, participants will be able to:

1. Understand the fundamentals of Retrieval-Augmented Generation (RAG) and its role in enhancing chatbot interactions.
2. Implement a user-friendly chatbot interface using Streamlit and integrate it with HuggingFace models.
3. Configure and manage context in chatbots using embeddings and similarity measures for relevant information retrieval.
4. Employ techniques for maintaining coherent conversation threads using Streamlit’s session state management.
5. Customize chatbots for specific domains and explore future trends and advancements in the field of NLP and conversational agents.

Prerequisites:

Participants should have:

• Basic understanding of Python programming.
• Familiarity with web application development concepts.
• Interest in NLP applications.
• Some experience with machine learning will be helpful

Note: Please enter your institutional/corporate email when registering.

Indicative Contents

1. Overview of HuggingFace and RAG. Introduce the audience to HuggingFace, a leading platform for natural language processing models, and Retrieval-Augmented Generation (RAG), an advanced method for improving chatbot responses by integrating external knowledge sources.
2. Understanding HuggingFace Hub. Discuss how HuggingFace Hub provides access to a wide range of pre-trained models, making it easier for developers to experiment and deploy NLP applications.
3. Integration with Streamlit. Explain the use of Streamlit for creating interactive web applications, and how it facilitates setting up a front-end for chatbot interaction with minimal effort.
4. Retrieval-Augmented Generation (RAG) Explained. Describe the concept of RAG, which combines retrieval of relevant information with the generative power of language models to enhance accuracy and relevance in responses.
5. Model Selection and Setup. Delve into the process of selecting an appropriate language model on HuggingFace and setting it up using the InferenceClient, highlighting considerations for choosing models based on task requirements.
6. Managing Context with RAG. Illustrate how RAG enhances chatbots by expanding their knowledge base using embeddings and similarity measures to retrieve contextually relevant information.
7. Embeddings for Contextual Search. Discuss the role of Sentence Transformers in generating embeddings, and how cosine similarity is employed to determine the relevance of retrieved information.
8. Building a Chatbot Interface. Showcase the potential of Streamlit’s user interface components, like chat inputs and message containers, allowing developers to create intuitive interactive experiences.
9. Handling Streaming Responses. Detail the method of handling and displaying streaming responses from language models using dynamic update features of Streamlit, enabling a seamless user experience.
10. Customization and Future Prospects. Conclude with advice on customizing chatbots for specific domains and discuss emerging trends in NLP and RAG that will further simplify the development of intelligent conversational agents.
11. Summary and Q&A. Wrap up the session by summarizing key takeaways and open the floor for questions, encouraging participants to explore HuggingFace and RAG for their projects.