Tips on What To Do With Your Language Model or API

Author(s): Louis Bouchard

Originally published on Towards AI.

Train, fine-tune, prompt, RAG… What to do?!

Do you ever question yourself if you should be training from scratch, fine-tuning, doing prompt engineering, or retrieving augmented generation (RAG)?

There are so many possibilities, but they each have a specific purpose and associated cost.

Here’s everything you need to know to enhance LLM performance, balancing quality, costs, and ease of use. U+2728U+1F680

Retrieval Augmented Generation (RAG) is now extremely popular. But what’s the difference between fine-tuning, simple “prompting”, or even training entirely from scratch? When should you use what?

Either launch a fast GPT-4 and explore prompt engineering and, once needed, try out fine-tuning for style-specific LLM adaptation without full retraining.

If you see lots of model hallucinations and/or misaligned output, try out RAG to enhance model accuracy and knowledge.

When it comes to fine-tuning, explore low-cost fine-tuning with LoRa and QLoRa. In the video and our free course (below), we cover large-scale model refinement and discuss training a model from scratch, including required datasets and resources.

This was a short overview of what you absolutely need to know… Learn more in this video that guides developers and AI enthusiasts on improving LLMs, offering methods for both minor and major advancements. Watch to refine LLM optimization skills:

P.S. If you found this post useful, we teach… Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI