Chapter 8 · Shaping models: tuning and alignment
Open-weight vs closed models
Open-weight models can be downloaded and run yourself; closed models are used through an API.
Preparing open-weight vs closed models
Chapter 8 · Shaping models: tuning and alignment
Open-weight models can be downloaded and run yourself; closed models are used through an API.
A file of weights is handed to you.
Your question goes in through the window.
Open-weight: you get the weights. Closed: you get answers.
You met weights before: the numbers a model learned. A model is mostly a very big file of them.
Pick a scenario or move the sliders. The bars show a simple teaching score, not real advice.
Leans open-weight. Keeping data on your own machines pulls you to open-weight.
What do you get with an open-weight model?
Which need most favours keeping data on your own machines?
A tiny weekend project with no hardware. What usually fits?