Caveat, this wasn’t Maya code, it was all our other giant piles of python code. But I’m operating under the assumption that the transition should be fairly similar.
Really the thing that 2to3 fails at, is distinguishing between when unadorned strings should be treated as unicode (or str now in python3) vs being converted into bytes. It errs on the side of turning everything into unicode, so if you’ve got string literals that need to remain as bytes, you need to manually convert them from "some_bytes" to b"some_bytes"