# What is LoRA? Definition & Technical Mechanics

> **Source:** [SPIDITS AI AI Glossary](https://spidits.com/ai-glossary)  
> **Category:** Model Training  
> **Canonical Citation:** [https://spidits.com/ai-glossary/lora](https://spidits.com/ai-glossary/lora)

## Definition
LoRA (Low-Rank Adaptation) is a parameter-efficient fine-tuning (PEFT) technique that freezes the pre-trained model weights and injects trainable rank decomposition matrices into each layer of the Transformer architecture, reducing training VRAM requirements.

## Architectural Mechanics & Deep Dive
LoRA (Low-Rank Adaptation) is a parameter-efficient fine-tuning (PEFT) technique that adapts large pre-trained models to specific tasks while frozen. It injects small trainable rank decomposition matrices into each layer of the Transformer, drastically reducing the number of parameters trained (by up to 99%) and lowering memory and training costs.

## Industry Applications & Real-World Use Cases
Resource-limited model fine-tuning, adapter model customization, and specialized behavior updates

## Frequently Asked Questions

### Why is LoRA so popular?
It reduces the number of trainable parameters by up to 99%, allowing developers to fine-tune 7B or 13B models on consumer-grade GPUs.

### What is a LoRA Adapter?
A tiny file containing the trained rank weights. These adapters can be dynamically swapped or merged onto the base model at runtime.

## Related Terminology
- [fine-tuning](https://spidits.com/ai-glossary/fine-tuning)
- [llm](https://spidits.com/ai-glossary/llm)
- [gpu](https://spidits.com/ai-glossary/gpu)

---
*From the SPIDITS AI Knowledge Base (336+ terms indexed). Explore full technical definitions: [https://spidits.com/ai-glossary](https://spidits.com/ai-glossary)*