Overview
Learn how to fine-tune the Falcon 7b Large Language Model on a custom dataset of chatbot customer support FAQs using QLoRA. Explore the process of loading the model, implementing a LoRA adapter, and conducting fine-tuning. Monitor training progress with TensorBoard and compare the performance of untrained and trained models by evaluating responses to various prompts. Gain insights into working with free-to-use LLMs for research and commercial purposes, and discover techniques for adapting powerful language models to specific tasks using limited computational resources.
Syllabus
- Introduction
- Text Tutorial on MLExpert.io
- Falcon LLM
- Google Colab Setup
- Dataset
- Load Falcon 7b and QLoRA Adapter
- Try the Model Before Training
- HuggingFace Dataset
- Training
- Save the Trained Model
- Load the Trained Model
- Evaluation
- Conclusion
Taught by
Venelin Valkov