[hetspookjee]

So much wh will one inference cost? I mean computing has come more efficient but I still struggle to find data on how much wh is used for some sentences of GPT3, for example.