pipette
ENEnglish

FRESCO: A Novel Consistency Control for Asynchronous Pipeline Parallel Training

Dongyeop Lee, Namhoon Lee

PreprintAfirmaciones fuertes, leer con cuidado

En palabras de los autores

Asynchronous pipeline parallelism can accelerate distributed model training, but suffers from instability caused by inconsistent gradient computations. We identify a fundamental source of this inconsistency: error signals are coupled to forward-time features, not just parameter states. This means that parameters may evolve as long as they preserve forward-equivalence. Leveraging this insight, we introduce FRESCO, which frames asynchronous updates as a constrained optimization problem: it finds the minimal update modification that acts trivially on the activation subspace. We demonstrate that FRESCO consistently outperforms representative asynchronous baselines in LLM pre-training, with the gains most pronounced under deep pipelines and high inconsistency levels: at pipeline depth 32, it reaches the same loss as the strongest baseline with 43% fewer iterations and 30% less wall-clock time, while requiring 8~921 less memory overhead. Moving beyond the current limitations, FRESCO establishes a new foundation for high-utilization pipeline parallel training, providing a scalable path for large-model development on diverse, resource-constrained systems.

Resultado principalEl resumen no menciona limitaciones.

Apareció: martes, 10 de noviembre. arXiv. Preprint, todavía sin revisión por pares.

Comentario de los autores: NeurIPS 2026