Metraipolis › archive › dc_62cbef875bd8

Nereus: Adaptive Parallelism for LLM Post-Training — arXiv:2609.34645

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.34645
License (as archived)
CC0-1.0
Archived
2026-10-07T21:02:46.727Z
SHA-256 of the archived text
bad153f2fcf33a03740758f946ded99e829783ded868adbaadbfd047fc1bfdde
Stable URL of this page
https://metraipolis.com/documento/dc_62cbef875bd8

How to cite

Nereus: Adaptive Parallelism for LLM Post-Training — arXiv:2609.34645 — arXiv (en), https://arxiv.org/abs/2609.34645 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_62cbef875bd8 (Metraipolis archive dc_62cbef875bd8, sha256 bad153f2fcf33a03740758f946ded99e829783ded868adbaadbfd047fc1bfdde)
BibTeX
@misc{dc62cbef875bd8,
  title = {Nereus: Adaptive Parallelism for LLM Post-Training — arXiv:2609.34645},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.34645},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_62cbef875bd8; sha256 bad153f2fcf33a03740758f946ded99e829783ded868adbaadbfd047fc1bfdde},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.34645. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Reinforcement learning (RL) post-training for large language models (LLMs) coordinates multiple models across generation, inference, and training on GPU clusters. Several factors may change during a run, including resource availability, sequence length, memory pressure, and stage bottlenecks. As a consequence, an execution plan that was initially suitable can then become slow or even infeasible over time. However, adapting a job whose models share GPUs entails significant challenges: deciding whether a new plan is worth the transition cost, reusing the job's distributed state, and coordinating GPU transfers across models and stages. Nereus targets these challenges as a cost-aware runtime that adapts RL post-training jobs into efficient execution plans. Its low-overhead controller selects a memory-feasible global plan and admits the transition using a cost model calibrated against the running job. To estimate and execute a transition, Nereus represents the distributed state of each replica of a model-stage (one model in one stage) as an Elastic Model Unit. It then employs a global transition graph to order the transformations and GPU transfers of these units. In a trace built from real data, online TP/PP adaptation reduces average step latency by 27.7% relative to the initial fixed TP/PP layout with DP scaling. In a 1,000-step run reaching 1,024 GPUs, six transitions consume 0.079% of total run time. Nereus improves end-to-end 8B PPO throughput by 2.14--7.27$\times$ over OpenRLHF and by 1.10--1.47$\times$ over Verl across diverse clusters.

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-29 in cs.AI).
Categorie arXiv: Distributed, Parallel, and Cluster Computing (cs.DC) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.34645 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-07T20:29:03.115Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-29?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).