IVF-Flat#

#include <raft/neighbors/ivf_flat.cuh>

namespace raft::neighbors::ivf_flat

template<typename T, typename IdxT>
auto extend(raft::resources const &handle, raft::device_matrix_view<const T, IdxT, row_major> new_vectors, std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices, const index<T, IdxT> &orig_index) -> index<T, IdxT>#

Build a new index containing the data of the original plus new extra vectors.

Implementation note: The new data is clustered according to existing kmeans clusters, then the cluster centers are adjusted to match the newly labeled data.

Usage example:

using namespace raft::neighbors;
ivf_flat::index_params index_params;
index_params.add_data_on_build = false;      // don't populate index on build
index_params.kmeans_trainset_fraction = 1.0; // use whole dataset for kmeans training
// train the index from a [N, D] dataset
auto index_empty = ivf_flat::build(handle, dataset, index_params, dataset);
// fill the index with the data
std::optional<raft::device_vector_view<const IdxT, IdxT>> no_op = std::nullopt;
auto index = ivf_flat::extend(handle, index_empty, no_op, dataset);

Template Parameters:
  • T – data element type

  • IdxT – type of the indices in the source dataset

Parameters:
  • handle[in]

  • new_vectors[in] raft::device_matrix_view to a row-major matrix [n_rows, index.dim()]

  • new_indices[in] optional raft::device_vector_view to a vector of indices [n_rows]. If the original index is empty (orig_index.size() == 0), you can pass std::nullopt here to imply a continuous range [0...n_rows).

  • orig_index[in] original index

Returns:

the constructed extended ivf-flat index

template<typename T, typename IdxT>
void extend(raft::resources const &handle, raft::device_matrix_view<const T, IdxT, row_major> new_vectors, std::optional<raft::device_vector_view<const IdxT, IdxT>> new_indices, index<T, IdxT> *index)#

Extend the index in-place with the new data.

Usage example:

using namespace raft::neighbors;
ivf_flat::index_params index_params;
index_params.add_data_on_build = false;      // don't populate index on build
index_params.kmeans_trainset_fraction = 1.0; // use whole dataset for kmeans training
// train the index from a [N, D] dataset
auto index_empty = ivf_flat::build(handle, index_params, dataset);
// fill the index with the data
std::optional<raft::device_vector_view<const IdxT, IdxT>> no_op = std::nullopt;
ivf_flat::extend(handle, dataset, no_opt, &index_empty);

Template Parameters:
  • T – data element type

  • IdxT – type of the indices in the source dataset

Parameters:
  • handle[in]

  • new_vectors[in] raft::device_matrix_view to a row-major matrix [n_rows, index.dim()]

  • new_indices[in] optional raft::device_vector_view to a vector of indices [n_rows]. If the original index is empty (orig_index.size() == 0), you can pass std::nullopt here to imply a continuous range [0...n_rows).

  • index[inout] pointer to index, to be overwritten in-place

template<typename T, typename IdxT, typename IvfSampleFilterT>
void search_with_filtering(raft::resources const &handle, const search_params &params, const index<T, IdxT> &index, raft::device_matrix_view<const T, IdxT, row_major> queries, raft::device_matrix_view<IdxT, IdxT, row_major> neighbors, raft::device_matrix_view<float, IdxT, row_major> distances, IvfSampleFilterT sample_filter = IvfSampleFilterT())#

Search ANN using the constructed index with the given filter.

See the ivf_flat::build documentation for a usage example.

Note, this function requires a temporary buffer to store intermediate results between cuda kernel calls, which may lead to undesirable allocations and slowdown. To alleviate the problem, you can pass a pool memory resource or a large enough pre-allocated memory resource to reduce or eliminate entirely allocations happening within search:

...
// use default search parameters
ivf_flat::search_params search_params;
filtering::none_ivf_sample_filter filter;
// Use the same allocator across multiple searches to reduce the number of
// cuda memory allocations
ivf_flat::search_with_filtering(
  handle, search_params, index, queries1, out_inds1, out_dists1, filter);
ivf_flat::search_with_filtering(
  handle, search_params, index, queries2, out_inds2, out_dists2, filter);
ivf_flat::search_with_filtering(
  handle, search_params, index, queries3, out_inds3, out_dists3, filter);
...

Template Parameters:
  • T – data element type

  • IdxT – type of the indices

  • IvfSampleFilterT – Device filter function, with the signature (uint32_t query_ix, uint32 cluster_ix, uint32_t sample_ix) -> bool or (uint32_t query_ix, uint32 sample_ix) -> bool

Parameters:
  • handle[in]

  • params[in] configure the search

  • index[in] ivf-flat constructed index

  • queries[in] a device pointer to a row-major matrix [n_queries, index->dim()]

  • neighbors[out] a device pointer to the indices of the neighbors in the source dataset [n_queries, k]

  • distances[out] a device pointer to the distances to the selected neighbors [n_queries, k]

  • sample_filter[in] a device filter function that greenlights samples for a given query

template<typename T, typename IdxT>
void search(raft::resources const &handle, const search_params &params, const index<T, IdxT> &index, raft::device_matrix_view<const T, IdxT, row_major> queries, raft::device_matrix_view<IdxT, IdxT, row_major> neighbors, raft::device_matrix_view<float, IdxT, row_major> distances)#

Search ANN using the constructed index.

See the ivf_flat::build documentation for a usage example.

Note, this function requires a temporary buffer to store intermediate results between cuda kernel calls, which may lead to undesirable allocations and slowdown. To alleviate the problem, you can pass a pool memory resource or a large enough pre-allocated memory resource to reduce or eliminate entirely allocations happening within search:

...
// use default search parameters
ivf_flat::search_params search_params;
// Use the same allocator across multiple searches to reduce the number of
// cuda memory allocations
ivf_flat::search(handle, search_params, index, queries1, out_inds1, out_dists1);
ivf_flat::search(handle, search_params, index, queries2, out_inds2, out_dists2);
ivf_flat::search(handle, search_params, index, queries3, out_inds3, out_dists3);
...

Template Parameters:
  • T – data element type

  • IdxT – type of the indices

Parameters:
  • handle[in]

  • params[in] configure the search

  • index[in] ivf-flat constructed index

  • queries[in] a device pointer to a row-major matrix [n_queries, index->dim()]

  • neighbors[out] a device pointer to the indices of the neighbors in the source dataset [n_queries, k]

  • distances[out] a device pointer to the distances to the selected neighbors [n_queries, k]

template<typename ValueT, typename IdxT, typename SizeT = uint32_t>
using list_data = ivf::list<list_spec, SizeT, ValueT, IdxT>#
static constexpr uint32_t kIndexGroupSize = 32#

Size of the interleaved group (see index::data description).

template<typename T, typename IdxT>
auto build(raft::resources const &handle, const index_params &params, raft::device_matrix_view<const T, IdxT, row_major> dataset) -> index<T, IdxT>#

Build the index from the dataset for efficient search.

NB: Currently, the following distance metrics are supported:

  • L2Expanded

  • L2Unexpanded

  • InnerProduct

Usage example:

using namespace raft::neighbors;
// use default index parameters
ivf_flat::index_params index_params;
// create and fill the index from a [N, D] dataset
auto index = ivf_flat::build(handle, dataset, index_params);
// use default search parameters
ivf_flat::search_params search_params;
// search K nearest neighbours for each of the N queries
ivf_flat::search(handle, search_params, index, queries, out_inds, out_dists);

Template Parameters:
  • T – data element type

  • IdxT – type of the indices in the source dataset

Parameters:
  • handle[in]

  • params[in] configure the index building

  • dataset[in] a device pointer to a row-major matrix [n_rows, dim]

Returns:

the constructed ivf-flat index

template<typename T, typename IdxT>
void build(raft::resources const &handle, const index_params &params, raft::device_matrix_view<const T, IdxT, row_major> dataset, raft::neighbors::ivf_flat::index<T, IdxT> &idx)#

Build the index from the dataset for efficient search.

NB: Currently, the following distance metrics are supported:

  • L2Expanded

  • L2Unexpanded

  • InnerProduct

Usage example:

using namespace raft::neighbors;
// use default index parameters
ivf_flat::index_params index_params;
// create and fill the index from a [N, D] dataset
ivf_flat::index<decltype(dataset::value_type), decltype(dataset::index_type)> index;
ivf_flat::build(handle, dataset, index_params, index);
// use default search parameters
ivf_flat::search_params search_params;
// search K nearest neighbours for each of the N queries
ivf_flat::search(handle, search_params, index, queries, out_inds, out_dists);

Template Parameters:
  • T – data element type

  • IdxT – type of the indices in the source dataset

Parameters:
struct index_params : public raft::neighbors::ann::index_params#
#include <ivf_flat_types.hpp>

Public Members

uint32_t n_lists = 1024#

The number of inverted lists (clusters)

uint32_t kmeans_n_iters = 20#

The number of iterations searching for kmeans centers (index building).

double kmeans_trainset_fraction = 0.5#

The fraction of data to use during iterative kmeans building.

bool adaptive_centers = false#

By default (adaptive_centers = false), the cluster centers are trained in ivf_flat::build, and never modified in ivf_flat::extend. As a result, you may need to retrain the index from scratch after invoking (ivf_flat::extend) a few times with new data, the distribution of which is no longer representative of the original training set.

The alternative behavior (adaptive_centers = true) is to update the cluster centers for new data when it is added. In this case, index.centers() are always exactly the centroids of the data in the corresponding clusters. The drawback of this behavior is that the centroids depend on the order of adding new data (through the classification of the added data); that is, index.centers() “drift” together with the changing distribution of the newly added data.

bool conservative_memory_allocation = false#

By default, the algorithm allocates more space than necessary for individual clusters (list_data). This allows to amortize the cost of memory allocation and reduce the number of data copies during repeated calls to extend (extending the database).

The alternative is the conservative allocation behavior; when enabled, the algorithm always allocates the minimum amount of memory required to store the given number of records. Set this flag to true if you prefer to use as little GPU memory for the database as possible.

struct search_params : public raft::neighbors::ann::search_params#
#include <ivf_flat_types.hpp>

Public Members

uint32_t n_probes = 20#

The number of clusters to search.

template<typename SizeT, typename ValueT, typename IdxT>
struct list_spec#
#include <ivf_flat_types.hpp>

Public Functions

inline constexpr auto make_list_extents(SizeT n_rows) const -> list_extents#

Determine the extents of an array enough to hold a given amount of data.

template<typename T, typename IdxT>
struct index : public raft::neighbors::ann::index#
#include <ivf_flat_types.hpp>

IVF-flat index.

Template Parameters:
  • T – data element type

  • IdxT – type of the indices in the source dataset

Public Functions

inline constexpr auto veclen() const noexcept -> uint32_t#

Vectorized load/store size in elements, determines the size of interleaved data chunks.

TODO: in theory, we can lift this to the template parameter and keep it at hardware maximum possible value by padding the dim of the data rapidsai/raft#711

inline constexpr auto metric() const noexcept -> raft::distance::DistanceType#

Distance metric used for clustering.

inline constexpr auto adaptive_centers() const noexcept -> bool#

Whether centers() change upon extending the index (ivf_pq::extend).

inline auto list_sizes() noexcept -> device_vector_view<uint32_t, uint32_t>#

Inverted list data [size, dim].

The data consists of the dataset rows, grouped by their labels (into clusters/lists). Within each list (cluster), the data is grouped into blocks of kIndexGroupSize interleaved vectors. Note, the total index length is slightly larger than the source dataset length, because each cluster is padded by kIndexGroupSize elements.

Interleaving pattern: within groups of kIndexGroupSize rows, the data is interleaved with the block size equal to veclen * sizeof(T). That is, a chunk of veclen consecutive components of one row is followed by a chunk of the same size of the next row, and so on.

Example: veclen = 2, dim = 6, kIndexGroupSize = 32, list_size = 31

x[ 0, 0], x[ 0, 1], x[ 1, 0], x[ 1, 1], ... x[14, 0], x[14, 1], x[15, 0], x[15, 1],
x[16, 0], x[16, 1], x[17, 0], x[17, 1], ... x[30, 0], x[30, 1],    -    ,    -    ,
x[ 0, 2], x[ 0, 3], x[ 1, 2], x[ 1, 3], ... x[14, 2], x[14, 3], x[15, 2], x[15, 3],
x[16, 2], x[16, 3], x[17, 2], x[17, 3], ... x[30, 2], x[30, 3],    -    ,    -    ,
x[ 0, 4], x[ 0, 5], x[ 1, 4], x[ 1, 5], ... x[14, 4], x[14, 5], x[15, 4], x[15, 5],
x[16, 4], x[16, 5], x[17, 4], x[17, 5], ... x[30, 4], x[30, 5],    -    ,    -    ,
Sizes of the lists (clusters) [n_lists] NB: This may differ from the actual list size if the shared lists have been extended by another index

inline auto centers() noexcept -> device_matrix_view<float, uint32_t, row_major>#

k-means cluster centers corresponding to the lists [n_lists, dim]

inline auto center_norms() noexcept -> std::optional<device_vector_view<float, uint32_t>>#

(Optional) Precomputed norms of the centers w.r.t. the chosen distance metric [n_lists].

NB: this may be empty if the index is empty or if the metric does not require the center norms calculation.

inline constexpr auto size() const noexcept -> IdxT#

Total length of the index.

inline constexpr auto dim() const noexcept -> uint32_t#

Dimensionality of the data.

inline constexpr auto n_lists() const noexcept -> uint32_t#

Number of clusters/inverted lists.

inline index(raft::resources const &res, raft::distance::DistanceType metric, uint32_t n_lists, bool adaptive_centers, bool conservative_memory_allocation, uint32_t dim)#

Construct an empty index. It needs to be trained and then populated.

inline index(raft::resources const &res, const index_params &params, uint32_t dim)#

Construct an empty index. It needs to be trained and then populated.

inline auto data_ptrs() noexcept -> device_vector_view<T*, uint32_t>#

Pointers to the inverted lists (clusters) data [n_lists].

inline auto inds_ptrs() noexcept -> device_vector_view<IdxT*, uint32_t>#

Pointers to the inverted lists (clusters) indices [n_lists].

inline constexpr auto conservative_memory_allocation() const noexcept -> bool#

Whether to use convervative memory allocation when extending the list (cluster) data (see index_params.conservative_memory_allocation).

inline void recompute_internal_state(raft::resources const &res)#

Update the state of the dependent index members.

inline auto lists() noexcept -> std::vector<std::shared_ptr<list_data<T, IdxT>>>&#

Lists’ data and indices.

Serializer Methods#

#include <raft/neighbors/ivf_flat_serialize.cuh>

namespace raft::neighbors::ivf_flat

template<typename T, typename IdxT>
void serialize(raft::resources const &handle, std::ostream &os, const index<T, IdxT> &index)#

Write the index to an output stream

Experimental, both the API and the serialization format are subject to change.

#include <raft/core/resources.hpp>

raft::resources handle;

// create an output stream
std::ostream os(std::cout.rdbuf());
// create an index with `auto index = ivf_flat::build(...);`
raft::serialize(handle, os, index);
Template Parameters:
  • T – data element type

  • IdxT – type of the indices

Parameters:
  • handle[in] the raft handle

  • os[in] output stream

  • index[in] IVF-Flat index

template<typename T, typename IdxT>
void serialize(raft::resources const &handle, const std::string &filename, const index<T, IdxT> &index)#

Save the index to file.

Experimental, both the API and the serialization format are subject to change.

#include <raft/core/resources.hpp>

raft::resources handle;

// create a string with a filepath
std::string filename("/path/to/index");
// create an index with `auto index = ivf_flat::build(...);`
raft::serialize(handle, filename, index);
Template Parameters:
  • T – data element type

  • IdxT – type of the indices

Parameters:
  • handle[in] the raft handle

  • filename[in] the file name for saving the index

  • index[in] IVF-Flat index

template<typename T, typename IdxT>
index<T, IdxT> deserialize(raft::resources const &handle, std::istream &is)

Load index from input stream

Experimental, both the API and the serialization format are subject to change.

#include <raft/core/resources.hpp>

raft::resources handle;

// create an input stream
std::istream is(std::cin.rdbuf());
using T    = float; // data element type
using IdxT = int; // type of the index
auto index = raft::deserialize<T, IdxT>(handle, is);
Template Parameters:
  • T – data element type

  • IdxT – type of the indices

Parameters:
  • handle[in] the raft handle

  • is[in] input stream

Returns:

raft::neighbors::ivf_flat::index<T, IdxT>

template<typename T, typename IdxT>
index<T, IdxT> deserialize(raft::resources const &handle, const std::string &filename)

Load index from file.

Experimental, both the API and the serialization format are subject to change.

#include <raft/core/resources.hpp>

raft::resources handle;

// create a string with a filepath
std::string filename("/path/to/index");
using T    = float; // data element type
using IdxT = int; // type of the index
auto index = raft::deserialize<T, IdxT>(handle, filename);
Template Parameters:
  • T – data element type

  • IdxT – type of the indices

Parameters:
  • handle[in] the raft handle

  • filename[in] the name of the file that stores the index

Returns:

raft::neighbors::ivf_flat::index<T, IdxT>

Helper Methods#

#include <raft/neighbors/ivf_flat_helpers.cuh>

namespace raft::neighbors::ivf_flat::helpers

template<typename T, typename IdxT>
void reset_index(const raft::resources &res, index<T, IdxT> *index)#

Public helper API to reset the data and indices ptrs, and the list sizes. Useful for externally modifying the index without going through the build stage. The data and indices of the IVF lists will be lost.

Usage example:

raft::resources res;
using namespace raft::neighbors;
// use default index parameters
ivf_flat::index_params index_params;
// initialize an empty index
ivf_flat::index<int64_t> index(res, index_params, D);
// reset the index's state and list sizes
ivf_flat::helpers::reset_index(res, &index);

Template Parameters:

IdxT

Parameters:
  • res[in] raft resource

  • index[inout] pointer to IVF-PQ index