Closed winglian closed 1 month ago
mistral 12b has a bunch of untrained reserved tokens that are useful for using as chatml tokens, this will allow us to use them without having inf as the grad norm.
inf
Weird I'm training with no inf/nan, and I started before this went into main
mistral 12b has a bunch of untrained reserved tokens that are useful for using as chatml tokens, this will allow us to use them without having
inf
as the grad norm.