Learn to fine-tune Google's Gemma 2B parameter model using Parameter Efficient Fine-tuning (PEFT) and LoRA in this 24-minute code walkthrough video. Master the process of running inference on pre-trained Gemma-2b, understand the motivation behind PEFT, and create a custom dataset derived from Databricks-dolly-15k for fine-tuning purposes. Explore essential tips for parameter optimization, supervised fine-tuning techniques, and interpret training visualizations through practical demonstrations in the HuggingFace ecosystem. Follow along with the provided Colab notebook to implement the concepts, from initial setup to final model evaluation. Benefit from comprehensive explanations of each step, including preliminaries, installation procedures, and detailed parameter configurations for optimal model performance.
Fine-tuning Gemma Models with PEFT and LoRA Using HuggingFace Datasets
Overview
Syllabus
- Intro
- Preliminaries & Installation
- Run inference on pre-trained Gemma-2b
- Motivation for Parameter Efficient Fine-tuning PEFT
- Create a custom dataset for finetuning
- Tips for fine-tuning parameters
- Supervised Fine-tuning
- Training visualization & Interpretations
- Conclusion
Taught by
AI Bites
Related Courses
-
Fine-Tuning LLM Models - Generative AI Course
-
Improving Accuracy of LLM Applications
-
Fine-Tuning LLMs with PEFT and LoRA
-
Parameter Efficient Fine-Tuning with Multiple LoRA Adapters for Large Language Models
-
Fine-tuning Large Language Models (LLMs) with Example Code
-
Understanding 4-bit Quantization and QLoRA - Memory Efficient Fine-tuning of LLMs
