Viewing profile — haoxiaoru

I've waited so long— four months

Recently, the Muon optimizer based on matrix orthogonalization has demonstrated strong results in training small-scale language models, but the scalability to larger models has not…