Faster Training of Diffusion Models and Improved Density Estimation via Parallel Score Matching - Inria - Institut national de recherche en sciences et technologies du numérique Accéder directement au contenu
Communication Dans Un Congrès Année : 2023

Faster Training of Diffusion Models and Improved Density Estimation via Parallel Score Matching

Résumé

In Diffusion Probabilistic Models (DPMs), the task of modeling the score evolution via a single time-dependent neural network necessitates extended training periods and may potentially impede modeling flexibility and capacity. To counteract these challenges, we propose leveraging the independence of learning tasks at different time points inherent to DPMs. More specifically, we partition the learning task by utilizing independent networks, each dedicated to learning the evolution of scores within a specific time sub-interval. Further, inspired by residual flows, we extend this strategy to its logical conclusion by employing separate networks to independently model the score at each individual time point. As empirically demonstrated on synthetic and image datasets, our approach not only significantly accelerates the training process by introducing an additional layer of parallelization atop data parallelization, but it also enhances density estimation performance when compared to the conventional training methodology for DPMs.Faster Training of Diffusion Models and Improved Density Estimation via Parallel Score Matching
Fichier principal
Vignette du fichier
Improving_Training_Speed_and_Density_Estimation_of_Diffusion_Models_via_Parallel_Score_Matching__HAL_.pdf.pdf (4.34 Mo) Télécharger le fichier
Improving_Training_Speed_and_Performance_of_Diffusion_Models_via_Parallel_Score_Matching__HAL_.pdf (3.31 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-04032669 , version 1 (21-03-2023)

Licence

Paternité

Identifiants

  • HAL Id : hal-04032669 , version 1

Citer

Etrit Haxholli, Marco Lorenzi. Faster Training of Diffusion Models and Improved Density Estimation via Parallel Score Matching. NeurIPS 2023 Workshop on Diffusion Models, Dec 2023, New Orleans, Louisiana, United States. ⟨hal-04032669⟩
136 Consultations
132 Téléchargements

Partager

Gmail Facebook X LinkedIn More