POET-X: Memory-efficient LLM Training by Scaling Orthogonal Transformation
2026-03-05 • Machine Learning
Machine LearningArtificial IntelligenceComputation and Language
AI summaryⓘ
The authors address the challenge of training large language models efficiently and stably. They improve on a previous method called POET by creating POET-X, which uses less memory and computation while keeping the original method's benefits. Their experiments show POET-X can train very large models on just one GPU, while older methods run out of memory. This means training big models can become more accessible and practical.
large language modelstraining stabilityorthogonal equivalence transformationPOETPOET-Xmatrix multiplicationmemory efficiencycomputational costGPUAdamW optimizer
Authors
Zeju Qiu, Lixin Liu, Adrian Weller, Han Shi, Weiyang Liu
Abstract
Efficient and stable training of large language models (LLMs) remains a core challenge in modern machine learning systems. To address this challenge, Reparameterized Orthogonal Equivalence Training (POET), a spectrum-preserving framework that optimizes each weight matrix through orthogonal equivalence transformation, has been proposed. Although POET provides strong training stability, its original implementation incurs high memory consumption and computational overhead due to intensive matrix multiplications. To overcome these limitations, we introduce POET-X, a scalable and memory-efficient variant that performs orthogonal equivalence transformations with significantly reduced computational cost. POET-X maintains the generalization and stability benefits of POET while achieving substantial improvements in throughput and memory efficiency. In our experiments, POET-X enables the pretraining of billion-parameter LLMs on a single Nvidia H100 GPU, and in contrast, standard optimizers such as AdamW run out of memory under the same settings.