every single one I put every single one through vit-l on my laptop because I didn't want to spend time figuring out how to optimize for a particular token because I already had tried for almost an hour
They are padded properly with the first token, then the body (the token we're encoding) then the last, then 0s filling out the 77 context length
