// ⚙ Fine-tuning & PEFT

LoRA (Low-Rank Adaptation)

Freezes the model and learns weight updates as small low-rank matrices.

LoRA 低秩适配 2021 Foundational core

IN PLAIN WORDS

Represent the update ΔW as the product of two thin matrices of small rank r, cutting trainable parameters by up to 10,000×; merge back for zero inference overhead. The most popular fine-tuning method.

Used in: Stable Diffusion style LoRAs, custom open models

TIMELINE · prerequisites → this → what builds on it

1901 Principal Com…2019 Adapters2021 LoRA (Low-Rank Ad…2022 Model Merging…2023 QLoRA2024 DoRA (Weight-…

01Original paper

LoRA: Low-Rank Adaptation of Large Language Models

Edward J. Hu, Yelong Shen, Phillip Wallis, et al. · 2021

arXiv ↗ ✓ Verified · arXiv API

02Where it sits

Learn the left column first; the right column builds on this one.

Learn first

LoRA (Low-Rank Adaptation)

Then you can learn

Left: prerequisites · right: builds on this · dashed: related. Full map →

03Learning resources

Official docs, explainers, courses and reference code. All links checked.

Official docs

huggingface.co ↗EN

Reference code

04Where the course teaches it

U11 Linear Algebra Vector spaces, matrix multiplication, eigenvalues and SVD — the language of embeddings, attention and LoRA. 15 h2 algorithms0 papers U50 Instruction Tuning & PEFT SFT, synthetic data, LoRA/QLoRA/DoRA, adapters and model merging. 10 h8 algorithms3 paperslesson ready

On learning paths:Post-training & Reasoning Models

For AI / data

Verification: verified, checked 2026-10-21.
JSON: /api/learn/algorithms.json (id: lora)