Library

NN.Library is the curated umbrella import for TorchLean's broad reusable library surface area.

It includes:

• the public user facade (NN.API.Public),
• subsystem entrypoints (NN.Entrypoint.*),
• the shared op-tagged IR through NN.Entrypoint.IR,
• ML-theory and tensor APIs,
• infoview tooling through NN.Entrypoint.Widgets,
• and proof-backed verification APIs through NN.Entrypoint.Verification.

It excludes:

• executables / demos that define main
• test suites and benchmark runners

Trust-boundary documentation lives in TRUST_BOUNDARIES.md, where it can cover both Lean declarations and external CUDA / Python / Julia / Arb producers without treating those prose notes are part of the Lean API.

If you’re writing ordinary model/training code, prefer import NN. If you want to make the broad umbrella explicit in downstream code, use import NN.Library. For lighter imports, use NN.Entrypoint.API, NN.Entrypoint.Tensor, NN.Entrypoint.IR, or another subsystem entrypoint.