DeepSpeed

9922270f - Further refactor deepspeed.moe.utils + deepspeed.moe.layer type hints (#5060)

Further refactor deepspeed.moe.utils + deepspeed.moe.layer type hints (#5060) When unpacking a `dict`, keys that appear after the unpacking can overwrite the keys of the unpacked `dict`, meaning we can avoid avoid the pattern of skipping certain keys; also use `defaultdict` to avoid having to do the boilerplate of assigning the elements of `group_moe`. More type hints and small stylistic changes to `deepspeed.moe.layer` --------- Co-authored-by: Michael Wyatt <mrwyattii@gmail.com> Co-authored-by: Michael Wyatt <michaelwyatt@microsoft.com>