Optimization Benchmark for Diffusion Models on Dynamical Systems

Fuente: arXiv
Saved in:
Bibliographic Details
Main Author: Schaipp, Fabian
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866911226636795904
author Schaipp, Fabian
author_facet Schaipp, Fabian
contents The training of diffusion models is often absent in the evaluation of new optimization techniques. In this work, we benchmark recent optimization algorithms for training a diffusion model for denoising flow trajectories. We observe that Muon and SOAP are highly efficient alternatives to AdamW (18% lower final loss). We also revisit several recent phenomena related to the training of models for text or image applications in the context of diffusion model training. This includes the impact of the learning-rate schedule on the training dynamics, and the performance gap between Adam and SGD.
format Preprint
id arxiv_https___arxiv_org_abs_2510_19376
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Optimization Benchmark for Diffusion Models on Dynamical Systems
Schaipp, Fabian
Machine Learning
Optimization and Control
The training of diffusion models is often absent in the evaluation of new optimization techniques. In this work, we benchmark recent optimization algorithms for training a diffusion model for denoising flow trajectories. We observe that Muon and SOAP are highly efficient alternatives to AdamW (18% lower final loss). We also revisit several recent phenomena related to the training of models for text or image applications in the context of diffusion model training. This includes the impact of the learning-rate schedule on the training dynamics, and the performance gap between Adam and SGD.
title Optimization Benchmark for Diffusion Models on Dynamical Systems
topic Machine Learning
Optimization and Control
url https://arxiv.org/abs/2510.19376