Hands-On AI: Building Your First Conversational AI Chatbot

Course details

Get up and running with designing, developing, and fine-tuning conversational chatbots using Hugging Face Transformers and PyTorch. Instructor Zhongyu Pan covers the essentials of creating a baseline AI-powered chatbot and improving it with fine-tuning on real-world dialogue datasets. Along the way, explore key concepts such as tokenization, model training, and evaluation techniques. By the end of this course, you’ll be equipped with highly marketable, in-demand skills to create a fully functional chatbot for real-world businesses.

Learning objectives
Build a baseline chatbot using Hugging Face Transformers and PyTorch.
Fine-tune a pretrained AI model with custom datasets to improve chatbot responses.
Deploy the chatbot as a web application using Gradio and Hugging Face Spaces.
Evaluate chatbot performance by comparing baseline and fine-tuned models.
Upgrade the chatbot with additional approaches.

Concepts

0. Introduction

01 - Hands-on conversational AI welcome
02 - Course roadmap - What you ll learn

1. Introduction to Conversational AI

03 - What is conversational AI
04 - Types of chatbots
05 - Applications of chatbots
06 - How to build chatbots

2. Setting Up the Development Environment

07 - Introduction to PyTorch and Hugging Face Transformers
08 - Setting up Google Colab and GPU for training
09 - Installing and importing required libraries

3. Building a Baseline Chatbot

10 - Introduction to pretrained conversational model - DialoGPT
11 - Loading DialoGPT
12 - Tokenization and preprocessing user input
13 - Generating responses with Transformers model

4. Launch Your First Chatbot Locally

14 - Introduction to Gradio
15 - Add CSS to your chatbot
16 - Setting up Gradio for your chatbot
17 - Launching and testing chatbot locally

5. Understanding Chatbot Fine-tuning

18 - Introduction to chatbot fine-tuning - What it is and why it matters
19 - Fine-tuning essentials - Datasets, overfitting, and generalization
20 - Choosing the right strategy - Fine-tuning vs. using pretrained models

6. Fine-Tuning the Chatbot for Better Conversations

21 - Choosing the right dataset
22 - Text data tokenization and preprocessing
23 - Fine-tuning with Trainer API
24 - Evaluate chatbot response quality

7. Further Upgrading Chatbot Responses

25 - Upgrade 1 - RAG (retrieval-augmented generation)
26 - Upgrade 2 - Improving response coherence and context awareness (and resources)
27 - Upgrade 3 - Handle uncertain responses with fallback mechanism

8. Launch Your Chatbot Online - Deploying on Hugging Face Spaces

28 - Creating a new space
29 - Upload your chatbot code
30 - Launching your chatbot online

Conclusion

31 - Course summary
32 - Next steps for learners