hAIgopala › AI terms › Rapid Fire (3 terms per short) › Day 24

LoRA / QLoRA

Model CustomisationChapter 7: Speaking in the School's VoiceTerm 119 of 253

Popular PEFT method adding small trainable adapters; QLoRA also quantises the model to save memory.

At Sunrise School

Gopal asks

Which method is popular?

Bit answers

LoRA adds small adapter layers. We keep one adapter for parents and another for students.

In the real world

Clip-on lens rather than a new camera.

Watch the short

Short coming soonSeason 5 · Day 24Follow @haigopala

Quick check

Quick check. Can fine-tuning make a model worse at old skills? Yes or no?

Taught together with

Catastrophic Forgetting · Reward Model

Practise it

Try this term and all 253 others in the free Model Training Arena.

Open the Arena →

Follow Bit daily: YouTube English · తెలుగు · हिंदी  |  Instagram English · తెలుగు · हिंदी