Search papers, labs, and topics across Lattice.
This paper introduces OmniOpt, a comprehensive survey and benchmarking framework for modern optimizers, addressing the fragmented landscape of over one hundred methods used in large-scale model training. By conceptualizing optimizer updates as structured transformations within a five-stage meta-pipeline and employing norm-constrained linear minimization oracles, the authors unify various optimizers into a dual-dimension taxonomy that categorizes them by mechanism and training objectives. The resulting cross-domain benchmark provides a systematic analysis of optimizer performance across diverse tasks, enabling researchers to make informed decisions based on explicit assumptions about mechanisms and objectives.
A unified framework reveals that most optimizers only engage a fraction of their potential, providing a roadmap for more effective model training.
Optimizer selection for large-scale model training has become a system-level design decision constrained jointly by compute, memory, tuning budget, and task diversity, yet the landscape of over one hundred methods remains fragmented. We therefore present OmniOpt, a unified survey and benchmark cookbook of optimizers for the research community. OmniOpt rests on four coupled components. First, we treat every optimizer update as a structured transformation through a five-stage meta-pipeline, and show that most methods engage only one or two of these stages. Second, we use norm-constrained linear minimization oracles (LMOs) to unify different optimizers. Third, these two views ground a dual-dimension taxonomy, one dimension assigning each method to a mechanism family and the other recording the measurable training objectives it aims to improve. Fourth, and at the core of this paper, we instantiate the full taxonomy in a unified cross-domain benchmark spanning representative optimizers, model scales, and training regimes from language model pretraining to image classification, systematically analyzing each method family across multiple effect objectives and laying out their trade-offs. OmniOpt thus supplies the research community with an operational coordinate system for selecting optimizers under explicit mechanism and objective assumptions, and charts a direction for the future development of the optimizer community.
