Struct WarpIndexConfig 

pub struct WarpIndexConfig {
    pub nbits: u8,
    pub num_centroids: usize,
    pub token_dim: usize,
    pub min_training_samples: Option<usize>,
    pub kmeans_iterations: usize,
}

Configuration for WARP index construction.

These parameters control the compression quality and index structure. The default values provide a good balance of memory efficiency and retrieval quality for most use cases.

Parameter Guidance

Corpus Size	nbits	num_centroids
< 100K docs	4	256
100K - 1M	2	1024
> 1M docs	2	4096

Fields

nbits: u8

Bits per dimension for residual quantization (2 or 4).

• 2-bit: 16× compression, ~3-5% MRR loss
• 4-bit: 8× compression, ~1-2% MRR loss

num_centroids: usize

Number of centroids for IVF clustering.

More centroids provide finer-grained partitioning but require more memory for centroid storage. Typical values: 256-4096.

token_dim: usize

Token embedding dimension (e.g., 128 for ColBERT).

min_training_samples: Option<usize>

Minimum training samples for codec training.

Should be at least 10 × num_centroids for stable clustering. If None, defaults to 10 × num_centroids.

kmeans_iterations: usize

K-means iterations for centroid training.

Implementations

impl WarpIndexConfig

pub fn new(nbits: u8, num_centroids: usize, token_dim: usize) -> Self

Create a new configuration with the specified parameters.

pub fn with_min_training_samples(self, samples: usize) -> Self

Set the minimum training samples.

pub fn with_kmeans_iterations(self, iterations: usize) -> Self

Set the k-means iterations.

pub fn effective_min_training_samples(&self) -> usize

Get the effective minimum training samples.

pub fn packed_residual_size(&self) -> usize

Calculate packed residual size in bytes.

pub fn validate(&self) -> Result<(), &'static str>

Validate the configuration.

Trait Implementations

impl Clone for WarpIndexConfig

fn clone(&self) -> WarpIndexConfig

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for WarpIndexConfig

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for WarpIndexConfig

fn default() -> Self

Returns the “default value” for a type.

impl<'de> Deserialize<'de> for WarpIndexConfig

fn deserialize<__D>(__deserializer: __D) -> Result<Self, __D::Error>

where __D: Deserializer<'de>,

Deserialize this value from the given Serde deserializer.

impl Serialize for WarpIndexConfig

fn serialize<__S>(&self, __serializer: __S) -> Result<__S::Ok, __S::Error>

where __S: Serializer,

Serialize this value into the given Serde serializer.