Faster Inference of Flow-Based Generative Models via Improved Data-Noise Coupling
This incremental improvement addresses faster inference for flow-based generative models, benefiting tasks like image and video generation.
The paper tackled the limitation of minibatch optimal transport in Conditional Flow Matching by introducing LOOM-CFM, which preserves and optimizes noise-data assignments across minibatches, resulting in consistent improvements in sampling speed-quality trade-offs across multiple datasets.
Conditional Flow Matching (CFM), a simulation-free method for training continuous normalizing flows, provides an efficient alternative to diffusion models for key tasks like image and video generation. The performance of CFM in solving these tasks depends on the way data is coupled with noise. A recent approach uses minibatch optimal transport (OT) to reassign noise-data pairs in each training step to streamline sampling trajectories and thus accelerate inference. However, its optimization is restricted to individual minibatches, limiting its effectiveness on large datasets. To address this shortcoming, we introduce LOOM-CFM (Looking Out Of Minibatch-CFM), a novel method to extend the scope of minibatch OT by preserving and optimizing these assignments across minibatches over training time. Our approach demonstrates consistent improvements in the sampling speed-quality trade-off across multiple datasets. LOOM-CFM also enhances distillation initialization and supports high-resolution synthesis in latent space training.