arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2602.19142cs.LGcs.AI
收藏

Celo2:迈向无学习优化的免费午餐

Celo2: Towards Learned Optimization Free Lunch

Mila(米拉);Concordia University(康考迪亚大学);Samsung AI Lab, Montreal(三星蒙特利尔人工智能实验室);Université de Montréal(蒙特利尔大学)

Abhinav Moudgil, Boris Knyazev, Eugene Belilovsky

AI总结:

Celo2通过简单架构和增强元训练,在极小计算量下实现了高效通用学习更新规则,稳定扩展至大规模预训练任务并兼容现代优化工具。

摘要:

已学习的优化器是替代手工设计的更新规则(如Adam)的强大替代品,但它们由于往往无法在训练分布之外进行元泛化以及元训练成本高而见到了有限的实用应用。例如,先前的工作VeLO将元训练扩展到4000个TPU月(约10倍于GPT-3的计算能力)以训练一个通用优化器,但其无法泛化到超过6亿参数的任务。在本工作中,我们提出了一个令人惊讶的发现:通过设计一个简单的归一化优化器架构并增强元训练,它变得可能在极小的VeLO计算量下(精确来说是4.5个GPU小时)元训练出一个高效的通用学习更新规则。我们的学习更新规则能够稳定地扩展到十亿级的预训练任务(GPT-3 XL 1.3B),其规模比元训练分布大了六个数量级。此外,它在各种分布外任务中表现强劲,并且与现代优化工具包兼容,包括正交化、输入-输出和隐藏权重的不同更新规则以及解耦的权重衰减。总之,这项工作为实际应用的可学习优化算法铺平了道路,解锁了更丰富的元训练和数据整理配方以进一步提高性能。

英文摘要:

Learned optimizers are powerful alternatives to hand-designed update rules like Adam, yet they have seen limited practical adoption since they often fail to meta-generalize beyond their training distribution and incur high meta-training cost. For instance, prior work, VeLO, scaled meta-training to 4,000 TPU months ($\sim$10$\times$ GPT-3 compute) to meta-train a general-purpose optimizer but it failed to generalize beyond 600M parameters tasks. In this work, we present a surprising finding: by crafting a simple normalized optimizer architecture and augmenting meta-training, it becomes feasible to meta-train a performant general-purpose learned update rule on a tiny fraction of VeLO compute, 4.5 GPU hours to be precise. Our learned update rule scales stably to a billion-scale pretraining task (GPT-3 XL 1.3B) which is six orders of magnitude larger than its meta-training distribution. Furthermore, it shows strong performance across diverse out-of-distribution tasks and is compatible with modern optimization harness that includes orthogonalization, distinct update rules for input-output and hidden weights, and decoupled weight decay. In all, this work paves the way for practically applicable learnable optimization algorithms, unlocking exploration of richer meta-training and data curation recipes to further improve performance.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑