Quickstart

Curated first-tour examples for TorchLean.

This umbrella is intentionally smaller than NN.Examples.Zoo. It teaches the primitives a new user needs before opening the model zoo:

• typed tensors and runtime scalar choices,
• editor widgets for inspecting tensors, floats, graphs, and logs,
• autograd helpers for gradients, Jacobians, Hessian-vector products, and detach,
• one compact end-to-end MLP training loop, and
• the proof/compile-time side of TorchLean's shape-indexed API.

Larger CNN, ResNet, data-loader, PyTorch interop, RL, and verification examples remain in their specialized folders. The point of Quickstart is a clean on-ramp, not another model zoo.