Open-source project
tensorflow/docs avatar
tensorflow/docs

tensorflow/docs: page sources, a git derived version, and no preview command

TensorFlow documentation

6,339 stars5,308 forksJupyter NotebookApache-2.0

At a glance

What is it?
tensorflow/docs is the source tree for the guide and tutorials on tensorflow.org, not the site itself: the README routes writers to three separate guides and sends every translation to another repository. The packaging script committed here is setup.py, and the version it reports is a timestamp of the tools/ directory rather than a released snapshot of the docs.
Who is it for?
Use tensorflow/docs if you intend to change a page on tensorflow.org or generate API reference pages from it, and read CONTRIBUTING.md plus the style guide before your first pull request, because the README itself carries no build command and no writing rules. Do not expect a local preview, an install command, or a release trail that matches the commit history: the only tag dates from 2023-05-25 while the tree was pushed on 2026-10-26.
Can I use it commercially?
Yes. Apache-2.0 is a permissive licence: you can use, modify and sell software built on it, as long as you keep its copyright and licence notices.
Is it still maintained?
Yes. The repository last received commits 11 days ago.
What is it written in?
Mainly Jupyter Notebook, according to GitHub's language statistics.

Answers come from the project's GitHub data, last synced on November 13, 2026, and from our analysis. They are not legal advice.

From daily snapshots

Star history

  • tensorflow/docs6,339 ★ · +9
Star history
Repository2026-09-132026-11-13
tensorflow/docs6.3k6.3k

9 Stars in the last 61 days, measured from the daily GitHub snapshot. Compare with other projects

Editorial analysis

Source files for tensorflow.org, not the site itself

This repository holds the source files for the guide and tutorials published on tensorflow.org, with the overview page as the entry point. It is not a mirror of the site: site/ carries the page sources and tools/ carries the generator that turns them into API reference pages.

The top level is short, and each entry says something about how the project is organized. setup.py packages the generator. MANIFEST.in decides which data files travel with it. CODEOWNERS routes review, AUTHORS records contributors, .gitattributes and .gitignore handle file handling, and CONTRIBUTING.md is where a new writer starts. The primary language here is Jupyter Notebook, so notebooks are a first class page format rather than an afterthought, which is also why a notebook reader sits in the required dependency list.

The package version is a git timestamp of tools/

Version numbers here describe a commit, not a documentation milestone. get_version() in setup.py asks git for the commit time of the tools directory, converts that timestamp to UTC, and prints year.month.day followed by the number of seconds elapsed since midnight, under a comment pointing at calver.org.

python
def get_version() -> str:
  ts = int(
      subprocess.check_output(['git', 'log', '-1', '--format=%ct', 'tools'])
      .decode('utf-8')
      .strip()
  )
  dt = datetime.datetime.utcfromtimestamp(ts)
  sec = 60 * 60 * dt.hour + 60 * dt.minute + dt.second

  # calver.org
  return f'{dt.year}.{dt.month}.{dt.day}.{sec}'

The single tagged release has that shape: 2023.5.24.56664, published on 2023-05-25, where 56664 is the second count of that day. Two limits follow from the mechanism. A build from an exported tree instead of a checkout has no history for git to answer, so setup.py stops inside subprocess.check_output before anything is installed. And two people working from different commits on the same day get two different versions, so the string says which commit of tools/ was used, not which pages were published.

Two dependency lists and a Python 3.9 floor

setup.py splits its dependencies in two. REQUIRED_PKGS holds astor, absl-py, jinja2, nbformat, protobuf>=3.12 and pyyaml: the notebook reader nbformat needs for the Jupyter pages, the template engine jinja2, the YAML parser pyyaml, and a protobuf floor of 3.12. VIS_REQUIRE holds numpy, PILLOW and webp.

Two things this pair of lists cannot do for you. Only protobuf carries a version constraint, and it is a floor with no ceiling, so nothing in the packaging metadata rules out a newer protobuf. And neither list includes the TensorFlow package itself, so installing both of them does not hand you an environment that can execute a tutorial. The python_requires value in the same file is '>=3.9', so an older interpreter is refused before any dependency is resolved.

No preview command and no console script in setup.py

Nothing in the README explains how to render the site. There is no install command, no build invocation, no port and no preview server named in it, and the setup() call in setup.py passes name, python_requires, version, description, long_description, author, author_email and url.

So the README cannot get you a running copy of tensorflow.org, and the dependency lists are a floor for the generator rather than a recipe for a local site. For contributing, the README points elsewhere. CONTRIBUTING.md sits at the repository root, the contributor guide lives at tensorflow.org/community/contribute/docs, and a separate style guide lives at tensorflow.org/community/contribute/docs_style. Anyone changing a page is expected to read all three before opening a pull request.

Docs issues are filed in the tensorflow/tensorflow tracker

A wrong sentence in a tutorial is not reported here. The README sends docs issues to the issue tracker in tensorflow/tensorflow, through the issue creation path that loads template=20-documentation-issue.md. The split matters for triage: commits in this repository touch Markdown and notebooks, while a page that renders wrongly on the live site can come from the site build or a translation, and neither of those lives in this tree.

The README also invites readers to join the TensorFlow documentation contributors on the TensorFlow Forum at discuss.tensorflow.org. Neither destination is versioned alongside the page sources, so whether a page has already been corrected has to be read from the commit history of the file itself rather than from a tracker whose numbering belongs to a different repository.

Translations live in docs-l10n on a best-effort basis

Translation is a separate project in a separate repository. The README sends community translations to tensorflow/docs-l10n and calls them contributed, reviewed and maintained by the community on a best-effort basis. A translator starts from site/<lang>/README.md, joins the mailing list for that language, and submits a pull request to the other repository.

Keeping the English sources in one place is the point of the split, but it also means the master branch and any translation can drift. A translation is a pull request against a moving base, and the project itself grades that work as best-effort. Reading a page in another language therefore tells you nothing about whether the English page behind it has changed, and nothing in this repository checks that a translated page matches the source it was translated from.

Pushed in 2026, tagged once in 2023

Commits and releases point in different directions here. The repository is not archived, and the last push was on 2026-10-26, so the page sources are still being worked on. The only tag, 2023.5.24.56664, was published on 2023-05-25 with the release note 'First release.'

That gap is the practical point for anyone installing the tensorflow-docs distribution. The tree moves; the published version has not moved since May 2023. Because the version number is derived from the tools/ directory, a fresh install can report a number that no release tag has ever used. Pinning on a tag pins a version string rather than a known good snapshot of the documentation, so compare the reported version against the tag list before you depend on it.

Cloning the sources and reading what setup.py requires

The README gives no install command, so there is no pip line to copy and nothing to paste. What it names is the project repository, and setup.py names the distribution, the interpreter floor and the two requirement lists. Getting the sources is a clone, and the version any build of that tree reports comes from a git log over tools/.

bash
git clone https://github.com/tensorflow/docs
cd docs
git log -1 --format=%ct tools

The last command is the one setup.py runs. It prints the commit time of the tools directory as a Unix timestamp, which is the input to the version string. The two requirement lists live beside it in the same file:

python
REQUIRED_PKGS = [
    'astor',
    'absl-py',
    'jinja2',
    'nbformat',
    'protobuf>=3.12',
    'pyyaml',
]

VIS_REQUIRE = [
    'numpy',
    'PILLOW',
    'webp',
]

Read them before installing anything; the python_requires value is '>=3.9'. The README does not say which published version of the site these sources correspond to, so that derived version string is the only marker available.

Editorial conclusion

Use tensorflow/docs if you intend to change a page on tensorflow.org or generate API reference pages from it, and read CONTRIBUTING.md plus the style guide before your first pull request, because the README itself carries no build command and no writing rules. Do not expect a local preview, an install command, or a release trail that matches the commit history: the only tag dates from 2023-05-25 while the tree was pushed on 2026-10-26. Before depending on the tensorflow-docs distribution, check whether the version it reports, which comes from a git log over tools/, lines up with a tag you can actually inspect.

Frequently asked questions

Is TensorFlow still relevant in 2026?

This repository answers only the documentation side of that question. It holds the source files for the guide and tutorials on tensorflow.org, it is not archived, and its last push was on 2026-10-26.

Is TensorFlow used anymore?

The README says nothing about who uses TensorFlow or how the site follows framework releases. It does show a documentation tree still being worked on, with a last push on 2026-10-26 and a single tagged release, 2023.5.24.56664, published on 2023-05-25.

What is TensorFlow used for?

That question is about the framework, and this repository only covers its documentation. The README points at the guide and tutorials on tensorflow.org and does not enumerate any use case for the framework itself.

Official sources

  1. License: Apache-2.0
  2. Project website
  3. README
  4. Releases
  5. tensorflow/docs on GitHub
Add this badge to your README

If you maintain this project, the badge below links readers to this analysis and shows its maintenance status from the daily GitHub snapshot. Paste the markdown into your README; add ?metric=license or ?metric=stars to the image URL for a different field.

Add this badge to your README

markdown
[![Hysen Labs](https://hysenlabs.com/badge/tensorflow-docs.svg)](https://hysenlabs.com/projects/tensorflow-docs)

Star history chart for your README

markdown
[![Star history](https://hysenlabs.com/chart/tensorflow-docs.svg)](https://hysenlabs.com/projects/tensorflow-docs)

Was this useful?