Planning to do this for the base model, at least at first. It's more of a learning thing for me, to help me perfect the context length extension approach so I can do it for our internal HyperWrite stuff… so I don't want to do too many variations.

→ View original post on X — @mattshumer_