ridge_superblock — n4m.compose.aom_superblock.ridge_superblock

Namespace: n4m.compose.aom_superblock · Fully-qualified: n4m.compose.aom_superblock.ridge_superblock · Catalog id: aom_pop.ridge_superblock

C ABI symbols (ABI 2.0): n · o · n · e

Python: from n4m.compose.aom_superblock import aom_ridge_superblock

Python-backed donor-style AOM Ridge superblock constrained to strict-linear single-operator AOM views. It builds operator outputs through native aom_preprocess, selects/fits Ridge alphas fold-locally through the native Ridge binding, applies train-fold centering and optional block RMS scaling to validation folds, and folds final superblock coefficients back to original-input input_coefficients plus intercept. It intentionally excludes donor branch_global, MKL/kernel, row-reference-dependent preprocessing and nonlinear AOM Ridge modes; native v1 builds in CUDA-enabled configurations but this is not yet a fused GPU superblock grinder.

Timing benchmark: benchmarks/cross_binding/bench_aom_ridge_superblock_timing.py

See also: methods index.