Resources
LoRA
- LoRA Paper - “LoRA: Low-Rank Adaptation of Large Language Models” (Hu et al., 2021)
- PEFT on GitHub - Hugging Face’s Parameter-Efficient Fine-Tuning library, supporting LoRA, QLoRA, and other adapters
QLoRA
- QLoRA Paper - “QLoRA: Efficient Finetuning of Quantized LLMs” (Dettmers et al., 2023)
- bitsandbytes on GitHub - Library providing 4-bit and 8-bit quantization support used by QLoRA
Fine-Tuning Frameworks
- TRL (Transformer Reinforcement Learning) - Hugging Face’s training library, including the SFTTrainer used in this module
- Unsloth on GitHub - Fast and memory-efficient QLoRA/LoRA training with optimized CUDA kernels
Weights & Biases
- Weights & Biases - ML experiment tracking and visualization platform used to monitor training runs
- W&B Quickstart - Getting started guide for logging metrics and comparing runs
- W&B Hugging Face Integration - Automatic integration with Hugging Face Trainer for zero-config logging
Hugging Face
- Hugging Face Hub - Central repository for sharing models, datasets, and demos
- Datasets Documentation - Library for loading, processing, and publishing datasets
- Model Cards Guide - Best practices for writing model cards, including recommended sections and metadata
Quantization and Local Deployment
- llama.cpp on GitHub - C/C++ library for GGUF quantization (
llama-quantize) and local inference - LM Studio - Desktop GUI for discovering, downloading, and chatting with quantized models from Hugging Face