ECHO-2: A Large-Scale Distributed Rollout Framework for Cost-Efficient Reinforcement Learning

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Song, Jingwei, Chen, Meng, Xiao, Jie, Ren, Qingnan, Huang, Jiaqi, Deng, Yangshen, Tong, Chris, Chen, Wanyi, Wang, Suli, Chen, Zhisheng, Bi, Ziqian, Lu, Shuo, Duan, Yiqun, Wang, Xu, Yu, Rymon, Ai, Lynn, Yang, Eric, Shi, Tianyu
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910256209068032
author Song, Jingwei
Chen, Meng
Xiao, Jie
Ren, Qingnan
Huang, Jiaqi
Deng, Yangshen
Tong, Chris
Chen, Wanyi
Wang, Suli
Chen, Zhisheng
Bi, Ziqian
Lu, Shuo
Duan, Yiqun
Wang, Xu
Yu, Rymon
Ai, Lynn
Yang, Eric
Shi, Tianyu
author_facet Song, Jingwei
Chen, Meng
Xiao, Jie
Ren, Qingnan
Huang, Jiaqi
Deng, Yangshen
Tong, Chris
Chen, Wanyi
Wang, Suli
Chen, Zhisheng
Bi, Ziqian
Lu, Shuo
Duan, Yiqun
Wang, Xu
Yu, Rymon
Ai, Lynn
Yang, Eric
Shi, Tianyu
contents Reinforcement learning (RL) is a critical stage in post-training large language models (LLMs), involving repeated interaction between rollout generation, reward evaluation, and centralized learning. Distributing rollout execution offers opportunities to leverage more cost-efficient inference resources, but introduces challenges in wide-area coordination and policy dissemination. We present ECHO-2, a distributed RL framework for post-training with remote inference workers and non-negligible dissemination latency. ECHO-2 combines centralized learning with distributed rollouts and treats bounded policy staleness as a user-controlled parameter, enabling rollout generation, dissemination, and training to overlap. We introduce an overlap-based capacity model that relates training time, dissemination latency, and rollout throughput, yielding a practical provisioning rule for sustaining learner utilization. To mitigate dissemination bottlenecks and lower cost, ECHO-2 employs peer-assisted pipelined broadcast and cost-aware activation of heterogeneous workers. Experiments on GRPO post-training of LLMs ranging from 4B to 32B parameters under real wide-area bandwidth regimes show that ECHO-2 significantly improves cost efficiency while preserving RL reward comparable to strong baselines.
format Preprint
id arxiv_https___arxiv_org_abs_2602_02192
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle ECHO-2: A Large-Scale Distributed Rollout Framework for Cost-Efficient Reinforcement Learning
Song, Jingwei
Chen, Meng
Xiao, Jie
Ren, Qingnan
Huang, Jiaqi
Deng, Yangshen
Tong, Chris
Chen, Wanyi
Wang, Suli
Chen, Zhisheng
Bi, Ziqian
Lu, Shuo
Duan, Yiqun
Wang, Xu
Yu, Rymon
Ai, Lynn
Yang, Eric
Shi, Tianyu
Machine Learning
Distributed, Parallel, and Cluster Computing
Reinforcement learning (RL) is a critical stage in post-training large language models (LLMs), involving repeated interaction between rollout generation, reward evaluation, and centralized learning. Distributing rollout execution offers opportunities to leverage more cost-efficient inference resources, but introduces challenges in wide-area coordination and policy dissemination. We present ECHO-2, a distributed RL framework for post-training with remote inference workers and non-negligible dissemination latency. ECHO-2 combines centralized learning with distributed rollouts and treats bounded policy staleness as a user-controlled parameter, enabling rollout generation, dissemination, and training to overlap. We introduce an overlap-based capacity model that relates training time, dissemination latency, and rollout throughput, yielding a practical provisioning rule for sustaining learner utilization. To mitigate dissemination bottlenecks and lower cost, ECHO-2 employs peer-assisted pipelined broadcast and cost-aware activation of heterogeneous workers. Experiments on GRPO post-training of LLMs ranging from 4B to 32B parameters under real wide-area bandwidth regimes show that ECHO-2 significantly improves cost efficiency while preserving RL reward comparable to strong baselines.
title ECHO-2: A Large-Scale Distributed Rollout Framework for Cost-Efficient Reinforcement Learning
topic Machine Learning
Distributed, Parallel, and Cluster Computing
url https://arxiv.org/abs/2602.02192