com.lancedb.lance.index.vector

Class PQBuildParams.Builder

java.lang.Object

com.lancedb.lance.index.vector.PQBuildParams.Builder

Enclosing class:

PQBuildParams

public static class PQBuildParams.Builder
extends Object

Constructor Summary

Constructors

Constructor and Description
Builder() Create a new builder for training a PQ model.

Method Summary

Modifier and Type	Method and Description
PQBuildParams	build()
PQBuildParams.Builder	setKmeansRedos(int kmeansRedos)
PQBuildParams.Builder	setMaxIters(int maxIters)
PQBuildParams.Builder	setNumBits(int numBits)
PQBuildParams.Builder	setNumSubVectors(int numSubVectors) The number of subvectors to divide the source vectors into.
PQBuildParams.Builder	setSampleRate(int sampleRate)

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait

Constructor Detail

Builder

public Builder()

Create a new builder for training a PQ model.

Method Detail

setNumSubVectors

public PQBuildParams.Builder setNumSubVectors(int numSubVectors)

The number of subvectors to divide the source vectors into. This must be a divisor of the vector dimension.

Parameters:

numSubVectors - the number of subvectors

Returns:

Builder

setNumBits

public PQBuildParams.Builder setNumBits(int numBits)

Parameters:

numBits - the number of bits to present one PQ centroid

Returns:

Builder

setMaxIters

public PQBuildParams.Builder setMaxIters(int maxIters)

Parameters:

maxIters - the max number of iterations for kmeans training.

Returns:

Builder

setKmeansRedos

public PQBuildParams.Builder setKmeansRedos(int kmeansRedos)

Parameters:

kmeansRedos - run kmeans `REDOS` times and take the best result. Default to 1

Returns:

Builder

setSampleRate

public PQBuildParams.Builder setSampleRate(int sampleRate)

Parameters:

sampleRate - sample rate to train PQ codebook

Returns:

Builder

build

public PQBuildParams build()