Class IndexIVFFlat

IndexIVFFlat Index. Inverted file with stored vectors.

Param

Param

The dimensionality of index.

Param

Number of clusters.

Param

Metric type (defaults to L2).

Hierarchy

Accessors

dims ids indexType isTrained metricArg metricType nprobe ntotal

Methods

add addWithIds dispose mergeFrom reconstruct reconstructBatch removeIds reset search toBuffer toIDMap2 train write fromBuffer fromFactory mergeOnDisk read

Accessors

dims

  • get dims(): number

  • Returns number

    The dimensionality of vectors.

ids

  • get ids(): BigInt[]

  • Vector identifiers, size ntotal.

    Returns BigInt[]

indexType

  • get indexType(): IndexType

  • Returns IndexType

    The type of index.

isTrained

  • get isTrained(): boolean

  • returns a boolean that indicates whether training is required.

    Returns boolean

    Whether training is required.

metricArg

  • get metricArg(): number

  • Returns number

    Argument of the metric type.

metricType

  • get metricType(): MetricType

  • Returns MetricType

    The metric of the index.

nprobe

  • get nprobe(): number

  • Cells to search.

    Returns number

  • set nprobe(value): void

  • Cells to search.

    Parameters

    • value: number

      The value to set.

    Returns void

ntotal

  • get ntotal(): number

  • Returns number

    The number of verctors currently indexed.

Methods

add

  • Add n vectors of dimension d to the index. Vectors are implicitly assigned labels ntotal .. ntotal + n - 1

    Parameters

    • x: number[]

      Input matrix, size n * d

    Returns void

addWithIds

  • Add n vectors of dimension d to the index using the provided labels.

    Parameters

    • x: number[]

      Input matrix, size n * d

    • y: (number | BigInt)[]

      Vector identifiers

    Returns void

  • Add n vectors of dimension d to the index with ID's.

    Parameters

    • x: number[]

      Input matrix, size n * d

    • ids: BigInt[]

      Vector identifiers

    Returns void

dispose

  • Free all resources associated with the index. Further calls to the index will throw.

    Returns void

mergeFrom

  • Merge the current index with another IndexIVFFlat instance.

    Parameters

    • otherIndex: IndexIVFFlat

      The other IndexIVFFlat instance to merge from.

    Returns void

reconstruct

  • Reconstruct desired vector from index. Will throw if not supported by the index type.

    Parameters

    • key: number | BigInt

      Key of vector to reconstruct.

    Returns number[]

    Reconstructed vector of length dims.

reconstructBatch

  • Reconstruct a batch of vectors from index. Will throw if not supported by the index type.

    Parameters

    • keys: (number | BigInt)[]

    Returns number[]

    Reconstructed vector of length dims.

removeIds

  • Remove IDs from the index.

    Parameters

    • ids: BigInt[]

      IDs to read.

    Returns number

    number of IDs removed.

reset

  • Reset the index, resulting in a ntotal of 0.

    Returns void

search

  • Query n vectors of dimension d to the index. return at most k vectors. If there are not enough results for a query, the result array is padded with -1s.

    Parameters

    • x: number[]

      Input vectors to search, size n * d.

    • k: number

      The number of nearest neighbors to search for.

    Returns SearchResult

    Output of the search result.

toBuffer

toIDMap2

train

  • Train n vectors of dimension d to the index. Vectors are implicitly assigned labels ntotal .. ntotal + n - 1

    Parameters

    • x: number[]

      Input matrix, size n * d

    Returns void

write

  • Write index to a file.

    Parameters

    • fname: string

      File path to write.

    Returns void

Static fromBuffer

Static fromFactory

  • Construct an index from factory descriptor.

    Parameters

    • dims: number

      Buffer to create index from.

    • descriptor: string

      Factory descriptor.

    • Optional metric: MetricType

      Metric type (defaults to L2).

    Returns Index

    The index read.

Static mergeOnDisk

  • Merge trained & untrained IVF indexes on disk.

    Parameters

    • mergePaths: string[]
    • outputIndexPath: string

      Path of final merged index file.

    • outputDataPath: string

      Path of final merged data file.

    Returns void

Static read

Settings

Member Visibility

Theme

On This Page

Generated using TypeDoc