LiMuon: Light and Fast Muon Optimizer for Large Models
Fuente:
arXiv
Saved in:
| Main Authors: | , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911730200739840 |
|---|---|
| author | Huang, Feihu Luo, Yuning Chen, Songcan |
| author_facet | Huang, Feihu Luo, Yuning Chen, Songcan |
| contents | Large models recently are widely applied in machine learning, so efficient training of large models has received widespread attention. More recently, the useful Muon optimizer is specifically designed for matrix-structured parameters of large models. Although some works have begun to study the Muon optimizer, the existing Muon and its variants still suffer from high sample complexity or high memory for large models. To fill this gap, we propose a light and fast Muon (LiMuon) optimizer for training large models, which builds on the momentum-based variance reduced technique and randomized Singular Value Decomposition (SVD). In particular, our LiMuon simultaneously has a lower memory and lower sample complexity than the Muon and its variants. Moreover, we prove that our LiMuon with lower memory has a lower sample complexity of $O(ε^{-3})$ for finding an $ε$-stationary solution of non-convex stochastic optimization under the generalized smooth condition. To further narrow practice and theory gap, we also prove that our LiMuon with Newton-Schulz steps has a lower sample complexity than the Muon with Newton-Schulz steps. Numerical experimental results on training Mamba-130M, Qwen2.5-0.5B and ViT models demonstrate effectiveness of our LiMuon. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2509_14562 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | LiMuon: Light and Fast Muon Optimizer for Large Models Huang, Feihu Luo, Yuning Chen, Songcan Machine Learning Optimization and Control Large models recently are widely applied in machine learning, so efficient training of large models has received widespread attention. More recently, the useful Muon optimizer is specifically designed for matrix-structured parameters of large models. Although some works have begun to study the Muon optimizer, the existing Muon and its variants still suffer from high sample complexity or high memory for large models. To fill this gap, we propose a light and fast Muon (LiMuon) optimizer for training large models, which builds on the momentum-based variance reduced technique and randomized Singular Value Decomposition (SVD). In particular, our LiMuon simultaneously has a lower memory and lower sample complexity than the Muon and its variants. Moreover, we prove that our LiMuon with lower memory has a lower sample complexity of $O(ε^{-3})$ for finding an $ε$-stationary solution of non-convex stochastic optimization under the generalized smooth condition. To further narrow practice and theory gap, we also prove that our LiMuon with Newton-Schulz steps has a lower sample complexity than the Muon with Newton-Schulz steps. Numerical experimental results on training Mamba-130M, Qwen2.5-0.5B and ViT models demonstrate effectiveness of our LiMuon. |
| title | LiMuon: Light and Fast Muon Optimizer for Large Models |
| topic | Machine Learning Optimization and Control |
| url | https://arxiv.org/abs/2509.14562 |