Mathematical Specifications

Import this file when you want TorchLean’s spec layer: shapes, tensors, layers, modules, and model constructors.

If you're writing specs or proofs, this is usually the right place to start. If you're trying to run models (autograd, training loops, import/export), use NN.API for the public application interface or NN.Runtime for the complete executable subsystem.

Structure:

• NN.Spec.Core for the core tensor/shape mathematics,
• NN.Spec.Layers.* for layer-level denotational semantics,
• NN.Spec.Module.* for PyTorch-style module wrappers over those specs,
• NN.Spec.Models.* for reusable model constructors,
• NN.Spec.Autograd.* / NN.Spec.Dynamics.* for auxiliary math-first interfaces.

The user-facing tensor constructors and notation live in NN.Tensor and are intentionally not re-exported here.