Expand description
Numeric / ML foundation for the legume ecosystem.
Modules:
leiden— vendored CWTS/10x Leiden port (seeNOTICE)matrix— matrix IO, KNN, layout, stats (exmatrix-util)param— parametric helpers (exmatrix-param)candle— candle training helpers (excandle-util; featurecandle)mcmc— MCMC diagnostics: effective sample size, split R̂ (featuremcmc)
Features: matrix and param work on nalgebra DMatrix alone.
tensor (default) adds candle Tensor support to the matrix code:
conversions, tensor IO, sampling and fused ops. ndarray (default) adds
ndarray Array2 support: IO, sampling, ndarray_stat, ndarray_gamma
and kNN from ndarray views. candle builds on tensor. A crate
that only needs DMatrix sets default-features = false and links
neither candle nor ndarray.
Re-exports§
pub use candle_core;