TensorPlay AI
Open navigation

Join the initiative

Students, developers, educators, and researchers can each change a different layer.

—Stars
—Contributors
—Commits (12 mo)
01

Students

Break things on purpose. File issues with what you expected versus what happened, add examples, and turn your study notes into docs.

  • Reproduce a bug and attach a minimal script
  • Add a worked example to the docs
  • Translate a tutorial
02

Developers

Every layer takes contributions: P10 kernels, TPX autograd, Stax graph passes, NN modules, and the dispatch plumbing between them.

  • Pick a good-first-issue kernel or fix
  • Add operator tests against the reference suite
  • Improve CUDA / ROCm / Vulkan coverage
03

Educators

The framework is a teaching instrument. Design labs that read tensors step by step, and share reproducible course material.

  • Build a lab from the trace-to-kernel path
  • Publish reproducible notebooks
  • Share workshop material
04

Researchers

Swap a layer, keep the rest readable: fusion experiments, new autograd modes, custom hardware kernels, static-graph capture.

  • Prototype fused kernels via Stax
  • Extend TPX with new gradient modes
  • Benchmark with the reproducible suites

How a contribution lands

  1. 01
    Pick an issue

    Comment to claim it; maintainers will confirm scope.

  2. 02
    Fork & branch

    Small, focused PRs land fastest.

  3. 03
    Prove it

    Tests plus benchmark output when performance-related.

  4. 04
    Review & merge

    CI gates must pass; expect review within days, not weeks.

Channels

Ask DeepWiki