PyTorch Lightning

license

Apache-2.0(OSI)

source

public(Lightning-AI/pytorch-lightning)

core features withheld

no — no enterprise or ee path in the repository root and no paid tier described in the README

PyTorch Lightning is released under a stock Apache-2.0 license, with the full source public in this repository. The README describes no enterprise or paid tier of the library itself; Lightning AI's hosted platform is a separate product sold alongside it, not a gated core.

• https://raw.githubusercontent.com/Lightning-AI/pytorch-lightning/master/LICENSE recorded 2026-10-06

  LICENSE body, read in full: the stock Apache License 2.0 text, byte-identical to the canonical template through END OF TERMS AND CONDITIONS and followed only by the standard appendix. No appended condition.

• https://raw.githubusercontent.com/Lightning-AI/pytorch-lightning/master/README.md recorded 2026-10-06

  README, read for the distribution and gating question: it documents pip-installing the library and running it locally, and describes no enterprise, paid or licence-gated build of the library. Lightning AI's hosted platform is a separate offering rather than a tier of this code.

Adoption is measured as PyPI downloads of the pytorch-lightning package; the umbrella lightning package, which redistributes the same code alongside Lightning Fabric, is not counted separately but points to comparable usage. GitHub stars are recorded only as corroboration, since a star is not an install.

• https://api.github.com/repos/Lightning-AI/pytorch-lightning recorded 2026-08-31

  Repo metadata for Lightning-AI/pytorch-lightning: stargazers_count 31313, archived false, pushed_at 2026-08-31.

• https://pypistats.org/api/packages/pytorch-lightning/recent recorded 2026-08-31

  last_month 12410527, last_week 2323920, last_day 238140 for the pytorch-lightning package.

PyTorch Lightning is a library built on PyTorch rather than a framework of its own: it structures and distributes a training loop, the same role Accelerate plays, and depends on PyTorch for the tensor engine and autograd underneath. Its breadth of accelerator and distributed-strategy support is what distinguishes it from Accelerate.

• https://raw.githubusercontent.com/Lightning-AI/pytorch-lightning/master/README.md recorded 2026-10-07

  The Key features section documents switching between CPU, GPU (CUDA and MPS), TPU, multi-GPU and multi-node execution with no code changes via the Fabric accelerator/devices/num_nodes arguments, and distributed strategies DDP, FSDP and DeepSpeed with mixed precision available out of the box.