File IndexHNSW.h

namespace faiss

Copyright (c) Facebook, Inc. and its affiliates.

This source code is licensed under the MIT license found in the LICENSE file in the root directory of this source tree.

Copyright (c) Facebook, Inc. and its affiliates.

This source code is licensed under the MIT license found in the LICENSE file in the root directory of this source tree. Implementation of k-means clustering with many variants.

Copyright (c) Facebook, Inc. and its affiliates.

This source code is licensed under the MIT license found in the LICENSE file in the root directory of this source tree. IDSelector is intended to define a subset of vectors to handle (for removal or as subset to search)

Copyright (c) Facebook, Inc. and its affiliates.

This source code is licensed under the MIT license found in the LICENSE file in the root directory of this source tree. PQ4 SIMD packing and accumulation functions

The basic kernel accumulates nq query vectors with bbs = nb * 2 * 16 vectors and produces an output matrix for that. It is interesting for nq * nb <= 4, otherwise register spilling becomes too large.

The implementation of these functions is spread over 3 cpp files to reduce parallel compile times. Templates are instantiated explicitly.

Copyright (c) Facebook, Inc. and its affiliates.

This source code is licensed under the MIT license found in the LICENSE file in the root directory of this source tree. This file contains callbacks for kernels that compute distances.

Throughout the library, vectors are provided as float * pointers. Most algorithms can be optimized when several vectors are processed (added/searched) together in a batch. In this case, they are passed in as a matrix. When n vectors of size d are provided as float * x, component j of vector i is

x[ i * d + j ]

where 0 <= i < n and 0 <= j < d. In other words, matrices are always compact. When specifying the size of the matrix, we call it an n*d matrix, which implies a row-major storage.

Copyright (c) Facebook, Inc. and its affiliates.

This source code is licensed under the MIT license found in the LICENSE file in the root directory of this source tree. I/O functions can read/write to a filename, a file handle or to an object that abstracts the medium.

The read functions return objects that should be deallocated with delete. All references within these objectes are owned by the object.

Copyright (c) Facebook, Inc. and its affiliates.

This source code is licensed under the MIT license found in the LICENSE file in the root directory of this source tree. Definition of inverted lists + a few common classes that implement the interface.

Copyright (c) Facebook, Inc. and its affiliates.

This source code is licensed under the MIT license found in the LICENSE file in the root directory of this source tree. Since IVF (inverted file) indexes are of so much use for large-scale use cases, we group a few functions related to them in this small library. Most functions work both on IndexIVFs and IndexIVFs embedded within an IndexPreTransform.

Copyright (c) Facebook, Inc. and its affiliates.

This source code is licensed under the MIT license found in the LICENSE file in the root directory of this source tree. In this file are the implementations of extra metrics beyond L2 and inner product

Copyright (c) Facebook, Inc. and its affiliates.

This source code is licensed under the MIT license found in the LICENSE file in the root directory of this source tree. Defines a few objects that apply transformations to a set of vectors Often these are pre-processing steps.

struct IndexHNSW : public faiss::Index
#include <IndexHNSW.h>

The HNSW index is a normal random-access index with a HNSW link structure built on top

Subclassed by faiss::IndexHNSW2Level, faiss::IndexHNSWCagra, faiss::IndexHNSWFlat, faiss::IndexHNSWPQ, faiss::IndexHNSWSQ

Public Types

typedef HNSW::storage_idx_t storage_idx_t

Public Functions

explicit IndexHNSW(int d = 0, int M = 32, MetricType metric = METRIC_L2)
explicit IndexHNSW(Index *storage, int M = 32)
~IndexHNSW() override
virtual void add(idx_t n, const float *x) override

Add n vectors of dimension d to the index.

Vectors are implicitly assigned labels ntotal .. ntotal + n - 1 This function slices the input vectors in chunks smaller than blocksize_add and calls add_core.

Parameters:
  • n – number of vectors

  • x – input matrix, size n * d

virtual void train(idx_t n, const float *x) override

Trains the storage if needed.

virtual void search(idx_t n, const float *x, idx_t k, float *distances, idx_t *labels, const SearchParameters *params = nullptr) const override

entry point for search

virtual void range_search(idx_t n, const float *x, float radius, RangeSearchResult *result, const SearchParameters *params = nullptr) const override

query n vectors of dimension d to the index.

return all vectors with distance < radius. Note that many indexes do not implement the range_search (only the k-NN search is mandatory).

Parameters:
  • n – number of vectors

  • x – input vectors to search, size n * d

  • radius – search radius

  • result – result table

virtual void reconstruct(idx_t key, float *recons) const override

Reconstruct a stored vector (or an approximation if lossy coding)

this function may not be defined for some indexes

Parameters:
  • key – id of the vector to reconstruct

  • recons – reconstucted vector (size d)

virtual void reset() override

removes all elements from the database.

void shrink_level_0_neighbors(int size)
void search_level_0(idx_t n, const float *x, idx_t k, const storage_idx_t *nearest, const float *nearest_d, float *distances, idx_t *labels, int nprobe = 1, int search_type = 1, const SearchParameters *params = nullptr) const

Perform search only on level 0, given the starting points for each vertex.

Parameters:

search_type – 1:perform one search per nprobe, 2: enqueue all entry points

void init_level_0_from_knngraph(int k, const float *D, const idx_t *I)

alternative graph building

void init_level_0_from_entry_points(int npt, const storage_idx_t *points, const storage_idx_t *nearests)

alternative graph building

void reorder_links()
void permute_entries(const idx_t *perm)

Public Members

HNSW hnsw
bool own_fields = false
Index *storage = nullptr
bool init_level0 = true
bool keep_max_size_level0 = false
struct IndexHNSWFlat : public faiss::IndexHNSW
#include <IndexHNSW.h>

Flat index topped with with a HNSW structure to access elements more efficiently.

Public Types

typedef HNSW::storage_idx_t storage_idx_t

Public Functions

IndexHNSWFlat()
IndexHNSWFlat(int d, int M, MetricType metric = METRIC_L2)
virtual void add(idx_t n, const float *x) override

Add n vectors of dimension d to the index.

Vectors are implicitly assigned labels ntotal .. ntotal + n - 1 This function slices the input vectors in chunks smaller than blocksize_add and calls add_core.

Parameters:
  • n – number of vectors

  • x – input matrix, size n * d

virtual void train(idx_t n, const float *x) override

Trains the storage if needed.

virtual void search(idx_t n, const float *x, idx_t k, float *distances, idx_t *labels, const SearchParameters *params = nullptr) const override

entry point for search

virtual void range_search(idx_t n, const float *x, float radius, RangeSearchResult *result, const SearchParameters *params = nullptr) const override

query n vectors of dimension d to the index.

return all vectors with distance < radius. Note that many indexes do not implement the range_search (only the k-NN search is mandatory).

Parameters:
  • n – number of vectors

  • x – input vectors to search, size n * d

  • radius – search radius

  • result – result table

virtual void reconstruct(idx_t key, float *recons) const override

Reconstruct a stored vector (or an approximation if lossy coding)

this function may not be defined for some indexes

Parameters:
  • key – id of the vector to reconstruct

  • recons – reconstucted vector (size d)

virtual void reset() override

removes all elements from the database.

void shrink_level_0_neighbors(int size)
void search_level_0(idx_t n, const float *x, idx_t k, const storage_idx_t *nearest, const float *nearest_d, float *distances, idx_t *labels, int nprobe = 1, int search_type = 1, const SearchParameters *params = nullptr) const

Perform search only on level 0, given the starting points for each vertex.

Parameters:

search_type – 1:perform one search per nprobe, 2: enqueue all entry points

void init_level_0_from_knngraph(int k, const float *D, const idx_t *I)

alternative graph building

void init_level_0_from_entry_points(int npt, const storage_idx_t *points, const storage_idx_t *nearests)

alternative graph building

void reorder_links()
void permute_entries(const idx_t *perm)

Public Members

HNSW hnsw
bool own_fields = false
Index *storage = nullptr
bool init_level0 = true
bool keep_max_size_level0 = false
struct IndexHNSWPQ : public faiss::IndexHNSW
#include <IndexHNSW.h>

PQ index topped with with a HNSW structure to access elements more efficiently.

Public Types

typedef HNSW::storage_idx_t storage_idx_t

Public Functions

IndexHNSWPQ()
IndexHNSWPQ(int d, int pq_m, int M, int pq_nbits = 8)
virtual void train(idx_t n, const float *x) override

Trains the storage if needed.

virtual void add(idx_t n, const float *x) override

Add n vectors of dimension d to the index.

Vectors are implicitly assigned labels ntotal .. ntotal + n - 1 This function slices the input vectors in chunks smaller than blocksize_add and calls add_core.

Parameters:
  • n – number of vectors

  • x – input matrix, size n * d

virtual void search(idx_t n, const float *x, idx_t k, float *distances, idx_t *labels, const SearchParameters *params = nullptr) const override

entry point for search

virtual void range_search(idx_t n, const float *x, float radius, RangeSearchResult *result, const SearchParameters *params = nullptr) const override

query n vectors of dimension d to the index.

return all vectors with distance < radius. Note that many indexes do not implement the range_search (only the k-NN search is mandatory).

Parameters:
  • n – number of vectors

  • x – input vectors to search, size n * d

  • radius – search radius

  • result – result table

virtual void reconstruct(idx_t key, float *recons) const override

Reconstruct a stored vector (or an approximation if lossy coding)

this function may not be defined for some indexes

Parameters:
  • key – id of the vector to reconstruct

  • recons – reconstucted vector (size d)

virtual void reset() override

removes all elements from the database.

void shrink_level_0_neighbors(int size)
void search_level_0(idx_t n, const float *x, idx_t k, const storage_idx_t *nearest, const float *nearest_d, float *distances, idx_t *labels, int nprobe = 1, int search_type = 1, const SearchParameters *params = nullptr) const

Perform search only on level 0, given the starting points for each vertex.

Parameters:

search_type – 1:perform one search per nprobe, 2: enqueue all entry points

void init_level_0_from_knngraph(int k, const float *D, const idx_t *I)

alternative graph building

void init_level_0_from_entry_points(int npt, const storage_idx_t *points, const storage_idx_t *nearests)

alternative graph building

void reorder_links()
void permute_entries(const idx_t *perm)

Public Members

HNSW hnsw
bool own_fields = false
Index *storage = nullptr
bool init_level0 = true
bool keep_max_size_level0 = false
struct IndexHNSWSQ : public faiss::IndexHNSW
#include <IndexHNSW.h>

SQ index topped with with a HNSW structure to access elements more efficiently.

Public Types

typedef HNSW::storage_idx_t storage_idx_t

Public Functions

IndexHNSWSQ()
IndexHNSWSQ(int d, ScalarQuantizer::QuantizerType qtype, int M, MetricType metric = METRIC_L2)
virtual void add(idx_t n, const float *x) override

Add n vectors of dimension d to the index.

Vectors are implicitly assigned labels ntotal .. ntotal + n - 1 This function slices the input vectors in chunks smaller than blocksize_add and calls add_core.

Parameters:
  • n – number of vectors

  • x – input matrix, size n * d

virtual void train(idx_t n, const float *x) override

Trains the storage if needed.

virtual void search(idx_t n, const float *x, idx_t k, float *distances, idx_t *labels, const SearchParameters *params = nullptr) const override

entry point for search

virtual void range_search(idx_t n, const float *x, float radius, RangeSearchResult *result, const SearchParameters *params = nullptr) const override

query n vectors of dimension d to the index.

return all vectors with distance < radius. Note that many indexes do not implement the range_search (only the k-NN search is mandatory).

Parameters:
  • n – number of vectors

  • x – input vectors to search, size n * d

  • radius – search radius

  • result – result table

virtual void reconstruct(idx_t key, float *recons) const override

Reconstruct a stored vector (or an approximation if lossy coding)

this function may not be defined for some indexes

Parameters:
  • key – id of the vector to reconstruct

  • recons – reconstucted vector (size d)

virtual void reset() override

removes all elements from the database.

void shrink_level_0_neighbors(int size)
void search_level_0(idx_t n, const float *x, idx_t k, const storage_idx_t *nearest, const float *nearest_d, float *distances, idx_t *labels, int nprobe = 1, int search_type = 1, const SearchParameters *params = nullptr) const

Perform search only on level 0, given the starting points for each vertex.

Parameters:

search_type – 1:perform one search per nprobe, 2: enqueue all entry points

void init_level_0_from_knngraph(int k, const float *D, const idx_t *I)

alternative graph building

void init_level_0_from_entry_points(int npt, const storage_idx_t *points, const storage_idx_t *nearests)

alternative graph building

void reorder_links()
void permute_entries(const idx_t *perm)

Public Members

HNSW hnsw
bool own_fields = false
Index *storage = nullptr
bool init_level0 = true
bool keep_max_size_level0 = false
struct IndexHNSW2Level : public faiss::IndexHNSW
#include <IndexHNSW.h>

2-level code structure with fast random access

Public Types

typedef HNSW::storage_idx_t storage_idx_t

Public Functions

IndexHNSW2Level()
IndexHNSW2Level(Index *quantizer, size_t nlist, int m_pq, int M)
void flip_to_ivf()
virtual void search(idx_t n, const float *x, idx_t k, float *distances, idx_t *labels, const SearchParameters *params = nullptr) const override

entry point for search

virtual void add(idx_t n, const float *x) override

Add n vectors of dimension d to the index.

Vectors are implicitly assigned labels ntotal .. ntotal + n - 1 This function slices the input vectors in chunks smaller than blocksize_add and calls add_core.

Parameters:
  • n – number of vectors

  • x – input matrix, size n * d

virtual void train(idx_t n, const float *x) override

Trains the storage if needed.

virtual void range_search(idx_t n, const float *x, float radius, RangeSearchResult *result, const SearchParameters *params = nullptr) const override

query n vectors of dimension d to the index.

return all vectors with distance < radius. Note that many indexes do not implement the range_search (only the k-NN search is mandatory).

Parameters:
  • n – number of vectors

  • x – input vectors to search, size n * d

  • radius – search radius

  • result – result table

virtual void reconstruct(idx_t key, float *recons) const override

Reconstruct a stored vector (or an approximation if lossy coding)

this function may not be defined for some indexes

Parameters:
  • key – id of the vector to reconstruct

  • recons – reconstucted vector (size d)

virtual void reset() override

removes all elements from the database.

void shrink_level_0_neighbors(int size)
void search_level_0(idx_t n, const float *x, idx_t k, const storage_idx_t *nearest, const float *nearest_d, float *distances, idx_t *labels, int nprobe = 1, int search_type = 1, const SearchParameters *params = nullptr) const

Perform search only on level 0, given the starting points for each vertex.

Parameters:

search_type – 1:perform one search per nprobe, 2: enqueue all entry points

void init_level_0_from_knngraph(int k, const float *D, const idx_t *I)

alternative graph building

void init_level_0_from_entry_points(int npt, const storage_idx_t *points, const storage_idx_t *nearests)

alternative graph building

void reorder_links()
void permute_entries(const idx_t *perm)

Public Members

HNSW hnsw
bool own_fields = false
Index *storage = nullptr
bool init_level0 = true
bool keep_max_size_level0 = false
struct IndexHNSWCagra : public faiss::IndexHNSW

Public Types

typedef HNSW::storage_idx_t storage_idx_t

Public Functions

IndexHNSWCagra()
IndexHNSWCagra(int d, int M, MetricType metric = METRIC_L2)
virtual void add(idx_t n, const float *x) override

Add n vectors of dimension d to the index.

Vectors are implicitly assigned labels ntotal .. ntotal + n - 1 This function slices the input vectors in chunks smaller than blocksize_add and calls add_core.

Parameters:
  • n – number of vectors

  • x – input matrix, size n * d

virtual void search(idx_t n, const float *x, idx_t k, float *distances, idx_t *labels, const SearchParameters *params = nullptr) const override

entry point for search

virtual void train(idx_t n, const float *x) override

Trains the storage if needed.

virtual void range_search(idx_t n, const float *x, float radius, RangeSearchResult *result, const SearchParameters *params = nullptr) const override

query n vectors of dimension d to the index.

return all vectors with distance < radius. Note that many indexes do not implement the range_search (only the k-NN search is mandatory).

Parameters:
  • n – number of vectors

  • x – input vectors to search, size n * d

  • radius – search radius

  • result – result table

virtual void reconstruct(idx_t key, float *recons) const override

Reconstruct a stored vector (or an approximation if lossy coding)

this function may not be defined for some indexes

Parameters:
  • key – id of the vector to reconstruct

  • recons – reconstucted vector (size d)

virtual void reset() override

removes all elements from the database.

void shrink_level_0_neighbors(int size)
void search_level_0(idx_t n, const float *x, idx_t k, const storage_idx_t *nearest, const float *nearest_d, float *distances, idx_t *labels, int nprobe = 1, int search_type = 1, const SearchParameters *params = nullptr) const

Perform search only on level 0, given the starting points for each vertex.

Parameters:

search_type – 1:perform one search per nprobe, 2: enqueue all entry points

void init_level_0_from_knngraph(int k, const float *D, const idx_t *I)

alternative graph building

void init_level_0_from_entry_points(int npt, const storage_idx_t *points, const storage_idx_t *nearests)

alternative graph building

void reorder_links()
void permute_entries(const idx_t *perm)

Public Members

bool base_level_only = false

When set to true, the index is immutable. This option is used to copy the knn graph from GpuIndexCagra to the base level of IndexHNSWCagra without adding upper levels. Doing so enables to search the HNSW index, but removes the ability to add vectors.

int num_base_level_search_entrypoints = 32

When base_level_only is set to True, the search function searches only the base level knn graph of the HNSW index. This parameter selects the entry point by randomly selecting some points and using the best one.

HNSW hnsw
bool own_fields = false
Index *storage = nullptr
bool init_level0 = true
bool keep_max_size_level0 = false