Training Deep Learning Models with Norm-Constrained LMOs
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2025
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866909640129773568 |
|---|---|
| author | Pethick, Thomas Xie, Wanyun Antonakopoulos, Kimon Zhu, Zhenyu Silveti-Falls, Antonio Cevher, Volkan |
| author_facet | Pethick, Thomas Xie, Wanyun Antonakopoulos, Kimon Zhu, Zhenyu Silveti-Falls, Antonio Cevher, Volkan |
| contents | In this work, we study optimization methods that leverage the linear minimization oracle (LMO) over a norm-ball. We propose a new stochastic family of algorithms that uses the LMO to adapt to the geometry of the problem and, perhaps surprisingly, show that they can be applied to unconstrained problems. The resulting update rule unifies several existing optimization methods under a single framework. Furthermore, we propose an explicit choice of norm for deep architectures, which, as a side benefit, leads to the transferability of hyperparameters across model sizes. Experimentally, we demonstrate significant speedups on nanoGPT training using our algorithm, Scion, without any reliance on Adam. The proposed method is memory-efficient, requiring only one set of model weights and one set of gradients, which can be stored in half-precision. The code is available at https://github.com/LIONS-EPFL/scion . |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2502_07529 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | Training Deep Learning Models with Norm-Constrained LMOs Pethick, Thomas Xie, Wanyun Antonakopoulos, Kimon Zhu, Zhenyu Silveti-Falls, Antonio Cevher, Volkan Machine Learning Optimization and Control In this work, we study optimization methods that leverage the linear minimization oracle (LMO) over a norm-ball. We propose a new stochastic family of algorithms that uses the LMO to adapt to the geometry of the problem and, perhaps surprisingly, show that they can be applied to unconstrained problems. The resulting update rule unifies several existing optimization methods under a single framework. Furthermore, we propose an explicit choice of norm for deep architectures, which, as a side benefit, leads to the transferability of hyperparameters across model sizes. Experimentally, we demonstrate significant speedups on nanoGPT training using our algorithm, Scion, without any reliance on Adam. The proposed method is memory-efficient, requiring only one set of model weights and one set of gradients, which can be stored in half-precision. The code is available at https://github.com/LIONS-EPFL/scion . |
| title | Training Deep Learning Models with Norm-Constrained LMOs |
| topic | Machine Learning Optimization and Control |
| url | https://arxiv.org/abs/2502.07529 |