Pré-Publication, Document De Travail (Preprint/Prepublication) Année : 2025

Improving Consistency Models with Generator-Augmented Flows

Thibaut Issenhuth
  • Fonction : Auteur
  • PersonId : 1392093
Ludovic dos Santos
Jean-Yves Franceschi
Alain Rakotomamonjy
  • Fonction : Auteur
  • PersonId : 1102415

Résumé

Consistency models imitate the multi-step sampling of score-based diffusion in a single forward pass of a neural network. They can be learned in two ways: consistency distillation and consistency training. The former relies on the true velocity field of the corresponding differential equation, approximated by a pre-trained neural network. In contrast, the latter uses a single-sample Monte Carlo estimate of this velocity field. The related estimation error induces a discrepancy between consistency distillation and training that, we show, still holds in the continuous-time limit. To alleviate this issue, we propose a novel flow that transports noisy data towards their corresponding outputs derived from a consistency model. We prove that this flow reduces the previously identified discrepancy and the noise-data transport cost. Consequently, our method not only accelerates consistency training convergence but also enhances its overall performance. The code is available at: \href{https://github.com/thibautissenhuth/consistency_GC}{ github.com/thibautissenhuth/consistency\_GC}.

Fichier principal
Vignette du fichier
GC_.pdf (3.13 Mo) Télécharger le fichier
Origine Fichiers produits par l'(les) auteur(s)
Licence

Dates et versions

hal-04611719 , version 1 (13-06-2024)
hal-04611719 , version 2 (14-10-2024)
hal-04611719 , version 3 (06-02-2025)
hal-04611719 , version 4 (02-07-2025)

Licence

Identifiants

  • HAL Id : hal-04611719 , version 3

Citer

Thibaut Issenhuth, Sangchul Lee, Ludovic dos Santos, Jean-Yves Franceschi, Chansoo Kim, et al.. Improving Consistency Models with Generator-Augmented Flows. 2025. ⟨hal-04611719v3⟩
362 Consultations
552 Téléchargements

Partager

  • More