core/vector_store/kmeans library
Lloyd's k-means clustering used by IVF-family indexes.
This is a straight port of FAISS' Clustering object at the level
of detail needed for coarse-quantizer training: k-means++ init,
fixed number of iterations, empty-cluster split, no minibatching.
Classes
- Kmeans
-
Simple Lloyd's k-means over
float32vectors with k-means++ init. - KmeansResult
- Result of a Kmeans.train call.