MuonBP: Faster Muon via Block-Periodic Orthogonalization

• URL: http://arxiv.org/abs/2510.16981v1
• Date: Sun, 19 Oct 2025 19:56:05 GMT
• Title: MuonBP: Faster Muon via Block-Periodic Orthogonalization
• Authors: Ahmed Khaled, Kaan Ozkara, Tao Yu, Mingyi Hong, Youngsuk Park,
• Abstract summary: We show how to adjust the learning rate from the baseline to MuonBP and give guarantees for this algorithm.<n>When training an 8B model with eight-way tensor tensor and ZeRO statewiseing, MuonBP achieves 8% Muon with no degradation in performance.
• Score: 24.232069944820513
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Gradient orthogonalization is a simple strategy that shows great utility in speeding up gradient descent. The Muon optimizer (Jordan, Jin, et al., 2024) combines gradient orthogonalization with first-order momentum and achieves significant improvement in data efficiency over Adam/AdamW (Loshchilov and Hutter, 2019) for language model training. However, when using model parallelism, gradient orthogonalization introduces additional overhead compared to coordinate-wise optimizers (such as AdamW) due to additional gather and scatter operations on gradient matrix shards from different devices. This additional communication can amount to a throughput hit of 5%-10% compared to Adam/AdamW. To remedy this, we propose Muon with Block-Periodic Orthogonalization (MuonBP), which applies orthogonalization independently to matrix shards on each device and periodically performs full orthogonalization to maintain training stability at scale. We show how to adjust the learning rate from the baseline to MuonBP and give convergence guarantees for this algorithm. Crucially, our theory dictates that we use two stepsizes: one for the blockwise orthogonalization steps, and one for the full orthogonalization steps. Our method is simple, requires minimal hyperparameter adjustments, and achieves competitive iteration complexity compared with baseline Muon while providing per-iteration throughput comparable to coordinate-wise methods such as AdamW. When training an 8B model with eight-way tensor parallelism and ZeRO optimizer state sharding, MuonBP achieves 8% throughput increase compared to Muon with no degradation in performance.

Related papers

• NorMuon: Making Muon more efficient and scalable [71.49702449498085]
  We propose NorMuon (Neuron-wise Normalized Muon) as a successor to Adam.<n>We show NorMuon consistently outperforms both Adam and Muon, achieving 21.74% better training efficiency than Adam and 11.31% improvement over Muon on 1.1 B pretraining setting.
  arXiv  Detail & Related papers  (2025-10-07T01:13:41Z)
• AuON: A Linear-time Alternative to Semi-Orthogonal Momentum Updates [0.0]
  We study the semi-orthogonal properties of momentum-based updates and develop a method to bound momentum updates under a spectral-norm trust region.<n>We propose AuON (Alternative Unit-norm momentum updates by Normalized nonlinear scaling), a linear-time that achieves strong performance without constructing semi-orthogonal matrices.<n>Our approach combines hyperbolic-cosine RMS scaling transformations with normalization, demonstrating both effectiveness and computational efficiency compared to Newton-Schulz methods.
  arXiv  Detail & Related papers  (2025-09-29T06:03:53Z)
• Effective Quantization of Muon Optimizer States [6.256712531304834]
  We introduce the 8-bit Muon using blockwise quantization, supporting both linear and dynamic schemes.<n>We demonstrate that 8-bit Muon maintains stability under both, while delivering $sim$74% reduction in memory footprint compared to full-precision Muon.<n>In extensive experiments, 8-bit Muon closely matches the performance of Muon while outperforming AdamW and 8-bit AdamW in pre-training a 1.6B model on 4B FineWeb tokens.
  arXiv  Detail & Related papers  (2025-09-27T04:31:11Z)
• AdaGrad Meets Muon: Adaptive Stepsizes for Orthogonal Updates [5.049533819651459]
  We propose a new adaptive update, AdaGO, which combines a norm-based update with aGrad-type step.<n>AdaGO preserves the orthogonality of the update, which can be interpreted as a spectral descent, while adapting the stepsizes to the optimization landscape by scaling the direction with accumulated past gradients.
  arXiv  Detail & Related papers  (2025-09-03T03:42:22Z)
• AdaMuon: Adaptive Muon Optimizer [11.281916426508216]
  AdaMuon combines element-wise adaptivity with orthogonal updates for large-scale neural network training.<n>AdaMuon maintains stability but can surpass Adam by more than 40% training efficiency in large-scale scenarios.
  arXiv  Detail & Related papers  (2025-07-15T05:49:37Z)
• Nesterov Method for Asynchronous Pipeline Parallel Optimization [59.79227116582264]
  We introduce a variant of Nesterov Accelerated Gradient (NAG) for asynchronous optimization in Pipeline Parallelism.<n>Specifically, we modify the look-ahead step in NAG to effectively address the staleness in gradients.<n>We theoretically prove that our approach converges at a sublinear rate in the presence of fixed delay in gradients.
  arXiv  Detail & Related papers  (2025-05-02T08:23:29Z)
• Muon is Scalable for LLM Training [50.68746986439438]
  We introduce Moonlight, a Mixture-of-Expert (MoE) model trained with 5.7T tokens using Muon.<n>Our model improves the current frontier, achieving better performance with much fewer training FLOPs compared to prior models.<n>We open-source our distributed Muon implementation that is memory optimal and communication efficient.
  arXiv  Detail & Related papers  (2025-02-24T09:12:29Z)
• MARS: Unleashing the Power of Variance Reduction for Training Large Models [56.67982828148859]
  We propose a unified training framework for deep neural networks.<n>We introduce three instances of MARS that leverage preconditioned gradient optimization.<n>Results indicate that the implementation of MARS consistently outperforms Adam.
  arXiv  Detail & Related papers  (2024-11-15T18:57:39Z)
• Breaking the Memory Barrier: Near Infinite Batch Size Scaling for Contrastive Loss [59.835032408496545]
  We propose a tile-based strategy that partitions the contrastive loss calculation into arbitrary small blocks. We also introduce a multi-level tiling strategy to leverage the hierarchical structure of distributed systems. Compared to SOTA memory-efficient solutions, it achieves a two-order-of-magnitude reduction in memory while maintaining comparable speed.
  arXiv  Detail & Related papers  (2024-10-22T17:59:30Z)
• 8-bit Optimizers via Block-wise Quantization [57.25800395197516]
  Statefuls maintain statistics over time, e.g., the exponentially smoothed sum (SGD with momentum) or squared sum (Adam) of past values. This state can be used to accelerate optimization compared to plain gradient descent but uses memory that might otherwise be allocated to model parameters. In this paper, we develop first gradients that use 8-bit statistics while maintaining the performance levels of using 32-bit gradient states.
  arXiv  Detail & Related papers  (2021-10-06T15:43:20Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.