hAIgopala › AI terms › Build Bit in 30 Days › Day 21

Fine-tuning

Model CustomisationChapter 7: Speaking in the School's VoiceTerm 117 of 253

Further training a pre-trained model on your own data to specialise its behaviour.

At Sunrise School

Gopal asks

We want replies to sound like our best coordinator.

Bit answers

Fine-tuning: train me on 2,000 of her best replies, and I adopt her warm, clear style.

In the real world

Train a model on 5,000 BASIS incident resolutions.

Watch the short

Short coming soonSeason 1 · Day 21Follow @haigopala

Quick check

Quick check. To update next year's fees: fine-tune, or RAG?

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