PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
December 3, 201916,138 citationsOpen Access

PyTorch: High-Performance Deep Learning Library with Imperative Style

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

View Full Paper
Ask AI
Bookmark
Share

Authors

APAdam PaszkeSGSam GrossFMFrancisco Massa

Discussion

Loading...

Member takes

Overview

Randomized trial demonstrates PyTorch's efficiency and performance in deep learning libraries, highlighting usability and speed compatibility.

Key Points

  • This research aims to explore the dual goals of usability and speed in deep learning frameworks, exemplified by PyTorch.
  • Detailed analysis of PyTorch's architecture and principles of implementation.
  • Performance demonstrations using common benchmarks to assess efficiency.
  • Evaluation of individual subsystems contributing to overall performance.
  • PyTorch combines a Pythonic programming style with high performance, achieving compelling efficiency.
  • Benchmarks indicate PyTorch's superior performance when compared to traditional deep learning frameworks.
  • Demonstrated effectiveness across various hardware accelerators, validating its usability and speed.

Cite This Study

Paszke et al. (2019) studied this question.

synapsesocial.com/papers/69ffcea5e4618ba4162d90b4https://doi.org/10.48550/arxiv.1912.01703
View Full Paper
Ask AI
Bookmark
Share