Transformer Models: An Introduction and Catalog

What I hoped for was a list of RAM requirements. Models are often presented as if it's irrelevant how many 80GB A100s you need. But in reality, trying out even inference often has a hard threshold based on the hardware you have at hand. And that hardware still is predominantly <= 24GB (3090 TI, A5000, A30).

Does anybody know of such a list? I.e. models with their respective minimum RAM requirements?

Hoped for the action figures, expected electronic component, got machine learning model. Box of chocolates.

Incomplete.

Most up-to-date:

https://lifearchitect.ai/models/

https://lifearchitect.ai/timeline/

This is a great quick reference.

Not enough to actually teach you what all these models do (that would be impossible), but I expect this will be great for times when I see a reference to a model I used to sort of understand and want a quick way to remember what it did.

Bookmarked.