One-step Flow Matching Generators
Keywords: one-step generator, text-to-image generation, flow matching
TL;DR: We introduce a state-of-the-art one-step text-to-image generator based on theory of flow generator matching.
Abstract: In the realm of Artificial Intelligence Generated Content (AIGC), flow-matching models have emerged as a powerhouse, achieving success due to their robust theoretical underpinnings and solid ability for large-scale generative modeling. These models have demonstrated state-of-the-art performance, but their brilliance comes at a cost. The process of sampling from these models is notoriously demanding on computational resources, as it necessitates the use of multi-step numerical ordinary differential equations (ODEs). Against this backdrop, this paper presents a novel solution with theoretical guarantees in the form of Flow Generator Matching (FGM), an innovative approach designed to accelerate the sampling of flow-matching models into a one-step generation, while maintaining the original performance. On the CIFAR10 unconditional generation benchmark, our one-step FGM model achieves a new record Fréchet Inception Distance (FID) score of 3.08 among all flow-matching-based models, outperforming flow matching models that use 50 generation steps.
Furthermore, we use the FGM to distill the Stable Diffusion 3, which is a leading text-to-image flow-matching model. The resulting model named the MM-DiT-FGM demonstrates outstanding industry-level performance as a novel transformer-based one-step text-to-image generator. When evaluated on GenEval benchmark, MM-DiT-FGM has delivered remarkable generating qualities, rivaling other multi-step models in light of the efficiency of a single generation step. We will release our one-step FGM text-to-image model with this paper.
Primary Area: generative models
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 8666
Loading