PyTorch: An Imperative Style, High-Performance Deep Learning Library

  • 类型:arxiv
  • 标识:1912.01703
  • 链接:https://arxiv.org/abs/1912.01703
  • 主题:evaluation
  • 主分类:engineering
  • 形态:method
  • 被引:54817
  • 被引来源:Semantic Scholar
  • S2被引:54817
  • OpenAlex被引:16173
  • 影响力被引:5900
  • TLDR:This paper details the principles that drove the implementation of PyTorch and how they are reflected in its architecture, and explains how the careful and pragmatic implementation of the key components of its runtime enables them to work together to achieve compelling performance.
  • OpenAlex ID:W2970971581
  • OpenAlex DOI:10.48550/arxiv.1912.01703
  • DOI:10.48550/arxiv.1912.01703
  • DOI来源:OpenAlex
  • 开放获取:green
  • 开放获取链接:https://arxiv.org/pdf/1912.01703
  • OpenAlex更新:2026-08-25
  • 待LLM分类:否
  • 成熟度:production
  • 场景:deep learning library、GPU acceleration、imperative graph
  • 标题中文:PyTorch:一种命令式风格的高性能深度学习库
  • TLDR中文:本文详细阐述了驱动 PyTorch 实现的原则及其在架构中的体现,并解释了 runtime 关键组件的精心且务实的实现如何使其协同工作以获得出色的性能。
  • 来源文件
  • [OpenAlex discover]
  • [S2 enrich]