Expand description
Data layer shared by training and inference.
The key design choice in this crate is that both dense and sparse storage are
exposed through the same TableAccess trait. Trainers therefore reason
about “rows, features, bins, and targets” instead of about concrete storage
formats. That keeps sampling views, inference-time reconstructed tables, and
Arrow-backed tables interoperable.
Structs§
- Dense
Table - Arrow-backed dense table for tabular regression/classification data.
- Sparse
Table - Arrow-backed sparse table specialized for binary feature matrices.
Enums§
Constants§
Traits§
- Table
Access - Common interface consumed by tree trainers and predictors.