API reference

This section provides detailed documentation for all the components of the library.

The API is organized into two main parts:

1. Available algorithms – Ready-to-use optimizers grouped by step-selection strategy (fixed step, line search, trust region).

2. Custom optimizers – A guide to assembling your own optimizers from the library’s building blocks: curvature estimators, line-search solvers, trust-region solvers, and linear solvers.

For a high-level overview of the library and its design, see the torch_numopt documentation.