1 paper
Aman Gupta, Rafael Celente, Abhishek Shivanna +7
The Muon optimizer, based on matrix orthogonalization, has recently shown faster convergence and better computational efficiency over AdamW in LLM pre-training. However, the memory…