4 ms·There are transformer-based approaches to vision models. It should work for those.by samus 3y agoThere are transformer-based approaches to vision models. It should work for those.