Module benchmark 

Benchmark utilities for ANN evaluation.

Provides metrics, dataset generation, and evaluation utilities for measuring ANN index quality across multiple dimensions:

• Accuracy: recall@k, precision@k, MRR
• Speed: QPS, latency percentiles
• Memory: bytes per vector, index overhead
• Scaling: performance vs dataset size, dimensionality

Standard Benchmark Datasets

Dataset	Size	Dim	Distance	Use Case
SIFT-1M	1M	128	L2	Image descriptors
GIST-1M	1M	960	L2	High-dimensional
GloVe-100	1.2M	100	Angular	Word embeddings
Fashion-MNIST	60K	784	L2	Small baseline

Reference: https://ann-benchmarks.com/

Re-exports

pub use datasets::compute_ground_truth;

pub use datasets::create_benchmark_dataset;

pub use datasets::Dataset;

pub use evaluation::cosine_distance;

pub use evaluation::evaluate;

pub use evaluation::generate_clustered_dataset;

pub use evaluation::generate_normalized_clustered_dataset;

pub use evaluation::generate_uniform_dataset;

pub use evaluation::l2_distance;

pub use evaluation::mrr;

pub use evaluation::normalize;

pub use evaluation::recall_at_k as eval_recall_at_k;

pub use evaluation::DistanceMetric;

pub use evaluation::EvalDataset;

pub use evaluation::EvalResults;

pub use memory::IndexMemoryStats;

pub use memory::MemoryTracker;

pub use metrics::precision_at_k;

pub use metrics::recall_at_k;

Modules

datasets

Synthetic and standard dataset generation for benchmarking.

evaluation

Evaluation metrics and infrastructure for ANN benchmarks.

memory

Memory tracking utilities for benchmarking.

metrics

Evaluation metrics for ANN quality.