[QUESTION] Muon args mismatch between Emerging Optimizers and Megatron-LM

[ShiftyBlock]

The OrthogonalizedOptimizer class in Emerging Optimizers uses

def __init__(
        ...
        *,
        nesterov: bool,
        ...
    ):

but currently the muon.py file uses use_nesterov instead of the nesterov OrthogonalizedOptimizer expects.

In a similar flavor, I also noticed that the Emerging Optimizers package currently uses tp_mode whereas currently MegatronLM uses mode.
I understand this integration is on the dev branch and the Emerging-Optimizers API is experimental/subject to change, so this may already be on your radar.