Will GPU be available during inference?

Will GPU be available during inference ?