Search papers, labs, and topics across Lattice.
This paper introduces Langevin Speculative Dynamics (LSD), a novel approach to accelerate molecular dynamics simulations by leveraging a draft model to propose rapid simulation steps while concurrently verifying them with a slower target model. By extending speculative sampling techniques from language and diffusion modeling to second-order Langevin dynamics, LSD achieves significant speedups of 3-9x across various systems without introducing relative error. The method's ability to generalize across different draft-target combinations and its theoretical and empirical validation demonstrate its potential to enhance throughput in molecular dynamics simulations.
Achieving 3-9x speedups in molecular dynamics simulations without compromising accuracy could revolutionize how we approach atomic system modeling.
Molecular dynamics (MD) is a key tool for simulating the dynamical behavior of atomic systems. However, MD is inherently serial, which makes it difficult to increase single-system throughput with concurrent compute. To address this, we introduce Langevin Speculative Dynamics (LSD), a distributed and model-agnostic speculative sampler for accelerating MD without adding relative error. Inspired by speculative methods in language and diffusion modeling, LSD uses a draft model to propose fast simulation steps and verifies them in parallel with a slower target model, applying a transport map from the draft to the target distribution. We extend speculative sampling to second-order Langevin dynamics, derive the achievable speedup as a function of physical parameters, show that LSD generalizes across different systems and draft-target combinations with a 3-9x speedup, and confirm theoretically and empirically that LSD samples trajectories from its target model distribution.
