LLM Fine-Tuning Course – From Supervised FT to RLHF, LoRA, and Multimodal
Sunny Savita
A 12-hour course on adapting pretrained language models to specific tasks. It begins with the theory: full versus partial fine-tuning, LoRA and QLoRA, instruction tuning and preference alignment with RLHF and DPO. Then it moves to code with Hugging Face, LLaMA-Factory and Unsloth, managed fine-tuning on the OpenAI API and Vertex AI, and finishes with embedding models and multimodal models such as Vision Transformers.
What you'll learn
- Parameter-efficient fine-tuning with LoRA and QLoRA
- Instruction fine-tuning with masked labels
- Aligning a model to preferences with DPO
- Training with LLaMA-Factory and Unsloth
- Fine-tuning jobs on the OpenAI API and Google Vertex AI
- Fine-tuning embedding models
