omnibias.torch.fastpath
PyTorch activation derivatives, trainable operator layers, network jets, and direct differential-operator contractions.
import omnibias.torch.fastpathModule contract & mathematical conventions
Closed-form fast-path kernels for the bias-collapse limit.
Per-activation kernel modules (eulerian for sigmoid,
hermite for Gaussian, legendre for tanh) export functions
of the form sigma_nth_derivative(z, n) -> sigma^(n)(z), evaluated in
closed form via the appropriate polynomial family.
The dispatch module exposes the literal multi-bias forward and
the collapsed-form forward used by OperatorMultiBiasUnit and
OperatorBlock.
Shared formulas do not imply universal floating-point identity. Verify dtype, shape, derivative-order support, and the stated numerical assumptions for the path you use.
Imports & aliases
These names share a definition with another public module. Follow a name for its full contract, parameters, and examples.
| Import from this module | Canonical reference |
|---|---|
bias_spread | omnibias.torch.fastpath.bias_spread |
is_collapsed | omnibias.torch.fastpath.is_collapsed |
multibias_collapsed_forward | omnibias.torch.fastpath.multibias_collapsed_forward |
multibias_forward | omnibias.torch.fastpath.multibias_forward |
multibias_literal_forward | omnibias.torch.fastpath.multibias_literal_forward |