Find your API
Start with the quantity you need. Choose a package, inspect its contracts, and follow a runnable example.
This reference follows the current RC/alpha workspace, with a recorded source revision. It includes public exports, module definitions, aliases, and project-defined class members. The installation guide explains version pins and compatibility with older consumers.
16 packages · reference at 9a2d73df0573
omnibias-binary
Binary and multilevel quantization with explicit hard-forward and surrogate-gradient contracts.
omnibias-boolean
Boolean truth tables, multilinear extensions, algebraic spectra, and finite-temperature gate systems.
omnibias-convex
Differentiable LP and QP solver paths, barrier methods, KKT sensitivities, and scoped optimality evidence.
omnibias-core
Coefficient recurrences, activation contracts, jet algebra, and verified arithmetic shared by the numerical backends.
omnibias-curvature
Scoped parameter-curvature formulas, Fisher and Gauss–Newton constructions, natural steps, and sharpness utilities.
omnibias-difference
Finite-difference stencils, confluence limits, polynomial sequences, and scoped derivative-extraction certificates.
omnibias-discrete
Shared discrete-problem contracts, annealed relaxations, hard decoders, and optimality-gap evidence.
omnibias-fields
Structured scalar, vector, and tensor fields with derivative caches, calculus operators, and coordinate conventions.
omnibias-graph
Spectral graph operators and differentiable relaxations for ordering, assignment, and graph-structured learning.
omnibias-jax
JAX activation derivatives, functional network jets, direct contractions, and parameter-differentiable numerical models.
omnibias-keras
Keras 3 operator layers and activation paths implemented with keras.ops for TensorFlow, JAX, or Torch.
omnibias-partition
Trainable soft partitions of unity, regional experts, temperature hardening, and scoped membership bounds.
omnibias-qcalculus
q-numbers, Jackson derivatives and integrals, q-series, and the ordinary-calculus limit as q approaches one.
omnibias-sos
Sum-of-squares constructions, polynomial positivity checks, and rigorous validation of proposed Gram matrices.
omnibias-struct
Differentiable dynamic programs for sequences and parses, structured distributions, and finite-temperature bounds.
omnibias-torch
PyTorch activation derivatives, trainable operator layers, network jets, and direct differential-operator contractions.
The source inventory covers the current public surface. Authored usage guidance and standalone examples are still expanding for advanced APIs. Read the source contract and linked tests before relying on an undocumented assumption; a signature alone does not establish supported shapes, dtypes, or numerical guarantees.
How to read the reference
Each package opens into focused module pages. Signatures, parameters, defaults, return annotations, explicit validation, and source links come from the pinned implementation. Public aliases link to one canonical reference rather than duplicating a contract.
Use the module contract for shape conventions and mathematical assumptions, then the individual entry for operation-specific details. Expand fields and methods for configuration and class behavior. Regression examples include their full source location and test command; fixtures belong to the pinned upstream checkout. See running reference examples.
Prefer the learning examples for complete, standalone scripts. Search accepts fully qualified names such as omnibias.torch.jet.mlp_jet.
Current substrate and external applications
The public substrate contains 16 distributions. PINN solver frontends, geometry, symbolic discovery, and other extracted applications have separate owners and compatibility requirements. The ecosystem guide preserves the distinction.