candding 0.1.0

Pure candle embeddings for Rust with no ONNX: dense, sparse, multi-vector and rerank models today, image models on the roadmap
Documentation 
1
2
3
4
5
6
7
8

//! BM42 (`Qdrant/all_miniLM_L6_v2_with_attentions`) over all-MiniLM-L6-v2: per-word CLS attention from the last
//! layer, max-folded by stem, rescored as `ln(1 + value) ^ 0.5` and hashed with `hash::bm25_hash`.

mod pooling;

pub mod head;

pub use head::Bm42Sparse;