LLM Fine-tuning

Parameter-efficient fine-tuning experiments for language models using QLoRA.

Project Structure

• Baseline/ - Foundation experiments with Phi-2 and Mistral-7B models
• code/ - CodeGen-2B fine-tuning for code generation tasks
• inference/ - Performance optimization techniques for inference

Quick Start

1. Navigate to desired experiment folder
2. Open corresponding Jupyter notebook
3. Follow setup instructions in each README

Requirements

• NVIDIA GPU (T4 recommended)
• Google Colab or local Jupyter environment
• Dependencies listed in requirements.txt

Hardware

Designed for consumer GPU constraints (15GB VRAM).