Generative Modeling with Phase Stochastic Bridges
AuthorsTianrong Chen, Jiatao Gu, Laurent Dinh, Evangelos A. Theodorou, Joshua Susskind, Shuangfei Zhai
Generative Modeling with Phase Stochastic Bridges
AuthorsTianrong Chen, Jiatao Gu, Laurent Dinh, Evangelos A. Theodorou, Joshua Susskind, Shuangfei Zhai
This paper introduces a novel generative modeling framework grounded in phase space dynamics, taking inspiration from the principles underlying Critically Damped Langevin Dynamics (CLD). Leveraging insights from stochastic optimal control, we construct a favorable path measure in the phase space that proves highly advantageous for generative sampling. A distinctive feature of our approach is the early-stage data prediction capability within the context of propagating generating Ordinary Differential Equations (ODEs) or Stochastic Differential Equations (SDEs) processes. This early prediction, enabled by the model’s unique structural characteristics, sets the stage for more efficient data generation, leveraging additional velocity information along the trajectory. This innovation has spurred the exploration of a novel avenue for mitigating sampling complexity by transitioning directly from noisy data to authentic images. Our model yields comparable results in image generation and notably outperforms baseline methods, particularly when faced with a limited Number of Function Evaluations (NFE). Furthermore, our approach rivals the performance of diffusion models equipped with efficient sampling techniques, underscoring its potential in the realm of generative modeling.
TADA: Improved Diffusion Sampling with Training-free Augmented Dynamics
September 22, 2025research area Computer Vision, research area Methods and Algorithmsconference NeurIPS
Diffusion models have demonstrated exceptional capabilities in generating high-fidelity images but typically suffer from inefficient sampling. Many solver designs and noise scheduling strategies have been proposed to dramatically improve sampling speeds. In this paper, we introduce a new sampling method that is up to 186% faster than the current state of the art solver for comparative FID on ImageNet512. This new sampling method is training-free…
KV-Runahead: Scalable Causal LLM Inference by Parallel Key-Value Cache Generation
May 14, 2024research area Methods and Algorithms, research area Speech and Natural Language Processingconference ICML
Large Language Model or LLM inference has two phases, the prompt (or prefill) phase to output the first token and the extension (or decoding) phase to the generate subsequent tokens. In this work, we propose an efficient parallelization scheme, KV-Runahead to accelerate the prompt phase. The key observation is that the extension phase generates tokens faster than the prompt phase because of key-value cache (KV-cache). Hence, KV-Runahead…