FastGRPO: Accelerating Policy Optimization via Concurrency-aware Speculative Decoding and Online Draft Learning
Group relative policy optimization (GRPO) has demonstrated significant potential in improving the reasoning capabilities of large language models (LLMs) via reinforcement learning. However, its practical deployment is impeded by an excessively slow training process, primarily attributed to the computationally intensive autoregressive generation of multiple responses per query, which makes the generation phase the primary performance bottleneck. Although speculative decoding presents a promising direction for acceleration, its direct application in GRPO achieves limited speedup under high-concurrency training conditions. To overcome this limitation, we propose a concurrency-aware speculative decoding framework that dynamically adjusts the drafting and verification strategy according to real-time concurrency levels, thereby maximizing the acceleration of the generation process. Furthermore, to address performance degradation arising from distributional drift between the evolving target model and the fixed draft model during training, we introduce an online draft learning mechanism that enables the draft model to continuously adapt using feedback signals from the target model. Experimental results across multiple mathematical reasoning datasets and models demonstrate that the proposed method achieves end-to-end speedups of 2.35x to 2.72x, significantly surpassing baseline approaches in efficiency. The code is available at https://github.com/yedaotian9/GRPO_speculative.
Code (0)
등록된 구현이 없습니다.
Tasks
Reinforcement LearningMathematical ReasoningSimilar Papers 제목 키워드 기반
Demystifying Parallel and Distributed Deep Learning: An In-Depth Concurrency Analysis
Deep Neural Networks (DNNs) are becoming an important tool in modern computing applications. Accelerating their training is a major challenge and techniques range from distributed algorithms to low-level circuit design. …
Deep LearningNeural Architecture SearchStochastic OptimizationA-3PO: Accelerating Asynchronous LLM Training with Staleness-aware Proximal Policy Approximation
Decoupled PPO has been a successful reinforcement learning (RL) algorithm to deal with the high data staleness under the asynchronous RL setting. Decoupled loss used in decoupled PPO improves coupled-loss style of algori…
Reinforcement LearningTailSieve: Partial-Rollout-Guided Tail Routing for LLM Rollouts
Large-scale rollouts have become a core component of modern LLM systems, spanning reinforcement learning (RL) post-training, on-policy distillation (OPD), and sampling-heavy evaluation pipelines. Unlike online serving, w…
Reinforcement LearningCharacterizing Concurrency Mechanisms for NVIDIA GPUs under Deep Learning Workloads
We investigate the performance of the concurrency mechanisms available on NVIDIA's new Ampere GPU microarchitecture under deep learning training and inference workloads. In contrast to previous studies that treat the GPU…
Deep LearningGPUSchedulingImplementing an intelligent version of the classical sliding-puzzle game for unix terminals using Golang's concurrency primitives
An intelligent version of the sliding-puzzle game is developed using the new Go programming language, which uses a concurrent version of the A* Informed Search Algorithm to power solver-bot that runs in the background. T…