arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2306.00577cs.LGcs.AI
收藏

TorchRL:一个用于 PyTorch 的数据驱动决策库

TorchRL: A data-driven decision-making library for PyTorch

Universitat Pompeu Fabra(庞培法布拉大学);Acellera(Acellera公司);University of Cambridge(剑桥大学);Meta AI;ICREA(加泰罗尼亚研究与高级研究院);PyTorch Team, Meta(Meta PyTorch团队)

Albert Bou, Matteo Bettini, Sebastian Dittert, Vikash Kumar, Shagun Sodhani, Xiaomeng Yang, Gianni De Fabritiis, Vincent Moens

AI总结:

为解决 PyTorch 缺乏原生决策控制库的问题,提出通用控制库 TorchRL,通过引入灵活原语 TensorDict 简化强化学习算法开发,实验证明其具备高可靠性、灵活性与计算效率,并在 GitHub 开源以促进社区协作。

摘要:

PyTorch 已崛起为顶级的机器学习框架,但它缺乏一个原生且全面的决策与控制任务库,无法满足处理复杂现实数据和环境的大型开发团队的需求。为解决此问题,我们提出了 TorchRL,这是一个面向 PyTorch 的通用控制库,提供集成良好且独立的组件。我们引入了一种新型且灵活的 PyTorch 原语 TensorDict,它有助于简化强化学习(RL)和控制众多分支的算法开发。我们详细描述了构建模块,并全面概述了该库在各个领域和任务中的应用。最后,我们通过实验证明了其可靠性和灵活性,并展示了对比基准以证明其计算效率。TorchRL 提供长期支持,已在 GitHub 上公开,以促进研究社区内更好的可复现性和协作。代码在 GitHub 上开源。

英文摘要:

PyTorch has ascended as a premier machine learning framework, yet it lacks a native and comprehensive library for decision and control tasks suitable for large development teams dealing with complex real-world data and environments. To address this issue, we propose TorchRL, a generalistic control library for PyTorch that provides well-integrated, yet standalone components. We introduce a new and flexible PyTorch primitive, the TensorDict, which facilitates streamlined algorithm development across the many branches of Reinforcement Learning (RL) and control. We provide a detailed description of the building blocks and an extensive overview of the library across domains and tasks. Finally, we experimentally demonstrate its reliability and flexibility and show comparative benchmarks to demonstrate its computational efficiency. TorchRL fosters long-term support and is publicly available on GitHub for greater reproducibility and collaboration within the research community. The code is open-sourced on GitHub.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑