Skip Navigation

What is better: higher quantiation or higher parameter count?

For example, does a 13B parameter model at 2_K quantiation perform worse than a 7B parameter model at 8bit or 16bit?

16

You're viewing a single thread.

16 comments
16 comments