3 ms·
Given how PALM2 had less params than the first one (540B) I think that makes sense.
by rileyphone 3y ago
Given how PALM2 had less params than the first one (540B) I think that makes sense.
- euclaise 3y agoCan't be less though, the latency difference between them implies that GPT-4 is significantly larger. This is unlike with PaLM, where latency decreased.