Open-source projects · Established

deepspeedai/DeepSpeed

Distributed deep learning optimization

  • ★ 43,272
  • 7 days +43
  • 30 days +169
  • Forks 5,059
  • Language：Python
  • License：Apache-2.0

Description on GitHub： DeepSpeed is a deep learning optimization library that makes distributed training and inference easy, efficient, and effective.

Highlights

  • DeepSpeed is a Python deep-learning optimization library for distributed training and inference.
  • Its README names ZeRO, ZeRO-Infinity, 3D parallelism, Ulysses sequence parallelism, and DeepSpeed-MoE among its system techniques.
  • The README says the library integrates with multiple open-source deep-learning frameworks.
  • The project uses the Apache-2.0 license.
  • TechSeeker measured 43,272 stars, with 43 added in the past week and 169 in the past month. Recent releases include several v0.19.x patch versions.

Our take

The latest listed releases are still in the 0.x series, so check upgrade notes for compatibility with existing training workflows. The supplied material does not specify supported accelerator and driver combinations or deployment requirements; verify these against the target cluster.

Star history

Star history：Stars grew from 0 to 43,272（2020-01-19 to 2026-10-25）
2020-01-19 Stars grew from 0 to 43,272 2026-10-25

Data updated: Oct 25, 2026

Releases

View on GitHub →

The summary and analysis were generated by AI. The original sources are authoritative. AI content policy