We released the entire training recipe for the Math model here (it's the same one we used for the v2):

→ View original post on X — @maximelabonne