CAGRA#

CAGRA is a graph-based nearest neighbors algorithm that was built from the ground up for GPU acceleration. CAGRA demonstrates state-of-the art index build and query performance for both small- and large-batch sized search.

#include <cuvs/neighbors/cagra.hpp>

namespace cuvs::neighbors::cagra

Index build parameters#

struct index_params : public cuvs::neighbors::index_params#

#include <cagra.hpp>

Public Members

size_t intermediate_graph_degree = 128#: Degree of input graph for pruning.

size_t graph_degree = 64#: Degree of output graph.

std::optional<cuvs::neighbors::vpq_params> compression = std::nullopt#: Specify compression parameters if compression is desired. If set, overrides the attach_dataset_on_build (and the compressed dataset is always added to the index).

std::variant<std::monostate, graph_build_params::ivf_pq_params, graph_build_params::nn_descent_params> graph_build_params#

Parameters for graph building.

Set ivf_pq_params or nn_descent_params to select the graph build algorithm and control their parameters. The default (std::monostate) is to use a heuristic to decide the algorithm and its parameters.

cagra::index_params params;
// 1. Choose IVF-PQ algorithm
params.graph_build_params = cagra::graph_build_params::ivf_pq_params(dataset.extent,
params.metric);

// 2. Choose NN Descent algorithm for kNN graph construction
params.graph_build_params =
cagra::graph_build_params::nn_descent_params(params.intermediate_graph_degree);

bool guarantee_connectivity = false#: Whether to use MST optimization to guarantee graph connectivity.

bool attach_dataset_on_build = true#

Whether to add the dataset content to the index, i.e.:

true means the index is filled with the dataset vectors and ready to search after calling build provided there is enough memory available.
false means build only builds the graph and the user is expected to update the dataset using cuvs::neighbors::cagra::update_dataset.

Regardless of the value of attach_dataset_on_build, the search graph is created using all the vectors in the dataset. Setting attach_dataset_on_build = false can be useful if the user needs to build only the search graph but does not intend to search it using CAGRA (e.g. search using another graph search algorithm), or if specific memory placement options need to be applied on the dataset before it is attached to the index using update_dataset. API.

auto dataset = raft::make_device_matrix<float, int64_t>(res, n_rows, n_cols);
// use default index_parameters
cagra::index_params index_params;
// update index_params to only build the CAGRA graph
index_params.attach_dataset_on_build = false;
auto index = cagra::build(res, index_params, dataset.view());
// assert that the dataset is not attached to the index
ASSERT(index.dataset().extent(0) == 0);
// update dataset
index.update_dataset(res, dataset.view());
// The index is now ready for search
cagra::search(res, search_params, index, queries, neighbors, distances);

Index search parameters#

enum class search_algo#

Values:

enumerator SINGLE_CTA#: For large batch sizes.

enumerator MULTI_CTA#: For small batch sizes.

enumerator MULTI_KERNEL#

enumerator AUTO#

enum class hash_mode#

Values:

enumerator HASH#

enumerator SMALL#

enumerator AUTO#

struct search_params : public cuvs::neighbors::search_params#

#include <cagra.hpp>

Public Members

size_t max_queries = 0#: Maximum number of queries to search at the same time (batch size). Auto select when 0.

size_t itopk_size = 64#

Number of intermediate search results retained during the search.

This is the main knob to adjust trade off between accuracy and search speed. Higher values improve the search accuracy.

size_t max_iterations = 0#: Upper limit of search iterations. Auto select when 0.

search_algo algo = search_algo::AUTO #: Which search implementation to use.

size_t team_size = 0#: Number of threads used to calculate a single distance. 4, 8, 16, or 32.

size_t search_width = 1#: Number of graph nodes to select as the starting point for the search in each iteration. aka search width?

size_t min_iterations = 0#: Lower limit of search iterations.

size_t thread_block_size = 0#: Thread block size. 0, 64, 128, 256, 512, 1024. Auto selection when 0.

hash_mode hashmap_mode = hash_mode::AUTO #: Hashmap type. Auto selection when AUTO.

size_t hashmap_min_bitlen = 0#: Lower limit of hashmap bit length. More than 8.

float hashmap_max_fill_rate = 0.5#: Upper limit of hashmap fill rate. More than 0.1, less than 0.9.

uint32_t num_random_samplings = 1#: Number of iterations of initial random seed node selection. 1 or more.

uint64_t rand_xor_mask = 0x128394#: Bit mask used for initial random seed node selection.

bool persistent = false#: Whether to use the persistent version of the kernel (only SINGLE_CTA is supported a.t.m.)

float persistent_lifetime = 2#: Persistent kernel: time in seconds before the kernel stops if no requests received.

float persistent_device_usage = 1.0#

Set the fraction of maximum grid size used by persistent kernel. Value 1.0 means the kernel grid size is maximum possible for the selected device. The value must be greater than 0.0 and not greater than 1.0.

One may need to run other kernels alongside this persistent kernel. This parameter can be used to reduce the grid size of the persistent kernel to leave a few SMs idle. Note: running any other work on GPU alongside with the persistent kernel makes the setup fragile.

Running another kernel in another thread usually works, but no progress guaranteed
Any CUDA allocations block the context (this issue may be obscured by using pools)
Memory copies to not-pinned host memory may block the context

Even when we know there are no other kernels working at the same time, setting kDeviceUsage to 1.0 surprisingly sometimes hurts performance. Proceed with care. If you suspect this is an issue, you can reduce this number to ~0.9 without a significant impact on the throughput.

Index extend parameters#

struct extend_params#

#include <cagra.hpp>

Public Members

uint32_t max_chunk_size = 0#: The additional dataset is divided into chunks and added to the graph. This is the knob to adjust the tradeoff between the recall and operation throughput. Large chunk sizes can result in high throughput, but use more working memory (O(max_chunk_size*degree^2)). This can also degrade recall because no edges are added between the nodes in the same chunk. Auto select when 0.

Index extend memory buffers#

Warning

doxygengroup: Cannot find group “cagra_cpp_extend_memory_buffers” in doxygen xml output for project “cuvs” from directory: ../../cpp/doxygen/_xml/

Index#

template<typename T, typename IdxT> struct index : public cuvs::neighbors::index#

#include <cagra.hpp>

CAGRA index.

The index stores the dataset and a kNN graph in device memory.

Template Parameters:

T – data element type
IdxT – type of the vector indices (represent dataset.extent(0))

Public Functions

inline constexpr auto metric() const noexcept -> cuvs::distance::DistanceType#: Distance metric used for clustering.

inline constexpr auto size() const noexcept -> IdxT #: Total length of the index (number of vectors).

inline constexpr auto dim() const noexcept -> uint32_t#: Dimensionality of the data.

inline constexpr auto graph_degree() const noexcept -> uint32_t#: Graph degree

inline auto data() const noexcept -> const cuvs::neighbors::dataset<int64_t>&#: Dataset [size, dim]

inline auto graph() const noexcept -> raft::device_matrix_view<const IdxT, int64_t, raft::row_major>#: neighborhood graph [size, graph-degree]

inline index(raft::resources const &res, cuvs::distance::DistanceType metric = cuvs::distance::DistanceType::L2Expanded)#: Construct an empty index.

template<typename data_accessor, typename graph_accessor> inline index(raft::resources const &res, cuvs::distance::DistanceType metric, raft::mdspan<const T, raft::matrix_extent<int64_t>, raft::row_major, data_accessor> dataset, raft::mdspan<const IdxT, raft::matrix_extent<int64_t>, raft::row_major, graph_accessor> knn_graph)#

Construct an index from dataset and knn_graph arrays

If the dataset and graph is already in GPU memory, then the index is just a thin wrapper around these that stores a non-owning a reference to the arrays.

The constructor also accepts host arrays. In that case they are copied to the device, and the device arrays will be owned by the index.

In case the dasates rows are not 16 bytes aligned, then we create a padded copy in device memory to ensure alignment for vectorized load.

Usage examples:

Cagra index is normally created by the cagra::build

using namespace raft::neighbors::experimental;
auto dataset = raft::make_host_matrix<float, int64_t>(n_rows, n_cols);
load_dataset(dataset.view());
// use default index parameters
cagra::index_params index_params;
// create and fill the index from a [N, D] dataset
auto index = cagra::build(res, index_params, dataset);
// use default search parameters
cagra::search_params search_params;
// search K nearest neighbours
auto neighbors = raft::make_device_matrix<uint32_t, int64_t>(res, n_queries, k);
auto distances = raft::make_device_matrix<float, int64_t>(res, n_queries, k);
cagra::search(res, search_params, index, queries, neighbors, distances);

In the above example, we have passed a host dataset to build. The returned index will own a device copy of the dataset and the knn_graph. In contrast, if we pass the dataset as a device_mdspan to build, then it will only store a reference to it.

Constructing index using existing knn-graph

using namespace raft::neighbors::experimental;

auto dataset = raft::make_device_matrix<float, int64_t>(res, n_rows, n_cols);
auto knn_graph = raft::make_device_matrix<uint32_n, int64_t>(res, n_rows, graph_degree);

// custom loading and graph creation
// load_dataset(dataset.view());
// create_knn_graph(knn_graph.view());

// Wrap the existing device arrays into an index structure
cagra::index<T, IdxT> index(res, metric, raft::make_const_mdspan(dataset.view()),
                            raft::make_const_mdspan(knn_graph.view()));

// Both knn_graph and dataset objects have to be in scope while the index is used because
// the index only stores a reference to these.
cagra::search(res, search_params, index, queries, neighbors, distances);

inline void update_dataset(raft::resources const &res, raft::device_matrix_view<const T, int64_t, raft::row_major> dataset)#

Replace the dataset with a new dataset.

If the new dataset rows are aligned on 16 bytes, then only a reference is stored to the dataset. It is the caller’s responsibility to ensure that dataset stays alive as long as the index. It is expected that the same set of vectors are used for update_dataset and index build.

inline void update_dataset(raft::resources const &res, raft::device_matrix_view<const T, int64_t, raft::layout_stride> dataset)#: Set the dataset reference explicitly to a device matrix view with padding.

inline void update_dataset(raft::resources const &res, raft::host_matrix_view<const T, int64_t, raft::row_major> dataset)#

Replace the dataset with a new dataset.

We create a copy of the dataset on the device. The index manages the lifetime of this copy. It is expected that the same set of vectors are used for update_dataset and index build.

template<typename DatasetT> inline auto update_dataset(raft::resources const &res, DatasetT &&dataset) -> std::enable_if_t<std::is_base_of_v<cuvs::neighbors::dataset<int64_t>, DatasetT>>#: Replace the dataset with a new dataset. It is expected that the same set of vectors are used for update_dataset and index build.

inline void update_graph(raft::resources const &res, raft::device_matrix_view<const IdxT, int64_t, raft::row_major> knn_graph)#

Replace the graph with a new graph.

Since the new graph is a device array, we store a reference to that, and it is the caller’s responsibility to ensure that knn_graph stays alive as long as the index.

inline void update_graph(raft::resources const &res, raft::host_matrix_view<const IdxT, int64_t, raft::row_major> knn_graph)#

Replace the graph with a new graph.

We create a copy of the graph on the device. The index manages the lifetime of this copy.

Index build#

auto build(raft::resources const &res, const cuvs::neighbors::cagra::index_params &params, raft::device_matrix_view<const float, int64_t, raft::row_major> dataset) -> cuvs::neighbors::cagra::index<float, uint32_t>#

Build the index from the dataset for efficient search.

The build consist of two steps: build an intermediate knn-graph, and optimize it to create the final graph. The index_params struct controls the node degree of these graphs.

The following distance metrics are supported:

L2
InnerProduct (currently only supported with IVF-PQ as the build algorithm)

Usage example:

using namespace cuvs::neighbors;
// use default index parameters
cagra::index_params index_params;
// create and fill the index from a [N, D] dataset
auto index = cagra::build(res, index_params, dataset);
// use default search parameters
cagra::search_params search_params;
// search K nearest neighbours
auto neighbors = raft::make_device_matrix<uint32_t>(res, n_queries, k);
auto distances = raft::make_device_matrix<float>(res, n_queries, k);
cagra::search(res, search_params, index, queries, neighbors, distances);

Parameters:

res – [in]
params – [in] parameters for building the index
dataset – [in] a matrix view (device) to a row-major matrix [n_rows, dim]

Returns:

the constructed cagra index

auto build(raft::resources const &res, const cuvs::neighbors::cagra::index_params &params, raft::host_matrix_view<const float, int64_t, raft::row_major> dataset) -> cuvs::neighbors::cagra::index<float, uint32_t>#

Build the index from the dataset for efficient search.

The build consist of two steps: build an intermediate knn-graph, and optimize it to create the final graph. The index_params struct controls the node degree of these graphs.

The following distance metrics are supported:

L2
InnerProduct (currently only supported with IVF-PQ as the build algorithm)

Usage example:

using namespace cuvs::neighbors;
// use default index parameters
cagra::index_params index_params;
// create and fill the index from a [N, D] dataset
auto index = cagra::build(res, index_params, dataset);
// use default search parameters
cagra::search_params search_params;
// search K nearest neighbours
auto neighbors = raft::make_device_matrix<uint32_t>(res, n_queries, k);
auto distances = raft::make_device_matrix<float>(res, n_queries, k);
cagra::search(res, search_params, index, queries, neighbors, distances);

Parameters:

res – [in]
params – [in] parameters for building the index
dataset – [in] a matrix view (host) to a row-major matrix [n_rows, dim]

Returns:

the constructed cagra index

auto build(raft::resources const &res, const cuvs::neighbors::cagra::index_params &params, raft::device_matrix_view<const half, int64_t, raft::row_major> dataset) -> cuvs::neighbors::cagra::index<half, uint32_t>#

Build the index from the dataset for efficient search.

The build consist of two steps: build an intermediate knn-graph, and optimize it to create the final graph. The index_params struct controls the node degree of these graphs.

The following distance metrics are supported:

L2
InnerProduct (currently only supported with IVF-PQ as the build algorithm)

Usage example:

using namespace cuvs::neighbors;
// use default index parameters
cagra::index_params index_params;
// create and fill the index from a [N, D] dataset
auto index = cagra::build(res, index_params, dataset);
// use default search parameters
cagra::search_params search_params;
// search K nearest neighbours
auto neighbors = raft::make_device_matrix<uint32_t>(res, n_queries, k);
auto distances = raft::make_device_matrix<float>(res, n_queries, k);
cagra::search(res, search_params, index, queries, neighbors, distances);

Parameters:

res – [in]
params – [in] parameters for building the index
dataset – [in] a matrix view (device) to a row-major matrix [n_rows, dim]

Returns:

the constructed cagra index

auto build(raft::resources const &res, const cuvs::neighbors::cagra::index_params &params, raft::host_matrix_view<const half, int64_t, raft::row_major> dataset) -> cuvs::neighbors::cagra::index<half, uint32_t>#

Build the index from the dataset for efficient search.

The build consist of two steps: build an intermediate knn-graph, and optimize it to create the final graph. The index_params struct controls the node degree of these graphs.

The following distance metrics are supported:

Usage example:

using namespace cuvs::neighbors;
// use default index parameters
cagra::index_params index_params;
// create and fill the index from a [N, D] dataset
auto index = cagra::build(res, index_params, dataset);
// use default search parameters
cagra::search_params search_params;
// search K nearest neighbours
auto neighbors = raft::make_device_matrix<uint32_t>(res, n_queries, k);
auto distances = raft::make_device_matrix<float>(res, n_queries, k);
cagra::search(res, search_params, index, queries, neighbors, distances);

Parameters:

res – [in]
params – [in] parameters for building the index
dataset – [in] a matrix view (host) to a row-major matrix [n_rows, dim]

Returns:

the constructed cagra index

auto build(raft::resources const &res, const cuvs::neighbors::cagra::index_params &params, raft::device_matrix_view<const int8_t, int64_t, raft::row_major> dataset) -> cuvs::neighbors::cagra::index<int8_t, uint32_t>#

Build the index from the dataset for efficient search.

The build consist of two steps: build an intermediate knn-graph, and optimize it to create the final graph. The index_params struct controls the node degree of these graphs.

The following distance metrics are supported:

Usage example:

using namespace cuvs::neighbors;
// use default index parameters
cagra::index_params index_params;
// create and fill the index from a [N, D] dataset
auto index = cagra::build(res, index_params, dataset);
// use default search parameters
cagra::search_params search_params;
// search K nearest neighbours
auto neighbors = raft::make_device_matrix<uint32_t>(res, n_queries, k);
auto distances = raft::make_device_matrix<float>(res, n_queries, k);
cagra::search(res, search_params, index, queries, neighbors, distances);

Parameters:

res – [in]
params – [in] parameters for building the index
dataset – [in] a matrix view (device) to a row-major matrix [n_rows, dim]

Returns:

the constructed cagra index

auto build(raft::resources const &res, const cuvs::neighbors::cagra::index_params &params, raft::host_matrix_view<const int8_t, int64_t, raft::row_major> dataset) -> cuvs::neighbors::cagra::index<int8_t, uint32_t>#

Build the index from the dataset for efficient search.

The build consist of two steps: build an intermediate knn-graph, and optimize it to create the final graph. The index_params struct controls the node degree of these graphs.

The following distance metrics are supported:

L2
InnerProduct (currently only supported with IVF-PQ as the build algorithm)

Usage example:

using namespace cuvs::neighbors;
// use default index parameters
cagra::index_params index_params;
// create and fill the index from a [N, D] dataset
auto index = cagra::build(res, index_params, dataset);
// use default search parameters
cagra::search_params search_params;
// search K nearest neighbours
auto neighbors = raft::make_device_matrix<uint32_t>(res, n_queries, k);
auto distances = raft::make_device_matrix<float>(res, n_queries, k);
cagra::search(res, search_params, index, queries, neighbors, distances);

Parameters:

res – [in]
params – [in] parameters for building the index
dataset – [in] a matrix view (host) to a row-major matrix [n_rows, dim]

Returns:

the constructed cagra index

auto build(raft::resources const &res, const cuvs::neighbors::cagra::index_params &params, raft::device_matrix_view<const uint8_t, int64_t, raft::row_major> dataset) -> cuvs::neighbors::cagra::index<uint8_t, uint32_t>#

Build the index from the dataset for efficient search.

The build consist of two steps: build an intermediate knn-graph, and optimize it to create the final graph. The index_params struct controls the node degree of these graphs.

The following distance metrics are supported:

L2
InnerProduct (currently only supported with IVF-PQ as the build algorithm)

Usage example:

using namespace cuvs::neighbors;
// use default index parameters
cagra::index_params index_params;
// create and fill the index from a [N, D] dataset
auto index = cagra::build(res, index_params, dataset);
// use default search parameters
cagra::search_params search_params;
// search K nearest neighbours
auto neighbors = raft::make_device_matrix<uint32_t>(res, n_queries, k);
auto distances = raft::make_device_matrix<float>(res, n_queries, k);
cagra::search(res, search_params, index, queries, neighbors, distances);

Parameters:

res – [in]
params – [in] parameters for building the index
dataset – [in] a matrix view (device) to a row-major matrix [n_rows, dim]

Returns:

the constructed cagra index

auto build(raft::resources const &res, const cuvs::neighbors::cagra::index_params &params, raft::host_matrix_view<const uint8_t, int64_t, raft::row_major> dataset) -> cuvs::neighbors::cagra::index<uint8_t, uint32_t>#

Build the index from the dataset for efficient search.

The build consist of two steps: build an intermediate knn-graph, and optimize it to create the final graph. The index_params struct controls the node degree of these graphs.

The following distance metrics are supported:

L2
InnerProduct (currently only supported with IVF-PQ as the build algorithm)

Usage example:

using namespace cuvs::neighbors;
// use default index parameters
cagra::index_params index_params;
// create and fill the index from a [N, D] dataset
auto index = cagra::build(res, index_params, dataset);
// use default search parameters
cagra::search_params search_params;
// search K nearest neighbours
auto neighbors = raft::make_device_matrix<uint32_t>(res, n_queries, k);
auto distances = raft::make_device_matrix<float>(res, n_queries, k);
cagra::search(res, search_params, index, queries, neighbors, distances);

Parameters:

res – [in]
params – [in] parameters for building the index
dataset – [in] a matrix view (host) to a row-major matrix [n_rows, dim]

Returns:

the constructed cagra index

Index search#

void search(raft::resources const &res, cuvs::neighbors::cagra::search_params const &params, const cuvs::neighbors::cagra::index<float, uint32_t> &index, raft::device_matrix_view<const float, int64_t, raft::row_major> queries, raft::device_matrix_view<uint32_t, int64_t, raft::row_major> neighbors, raft::device_matrix_view<float, int64_t, raft::row_major> distances, const cuvs::neighbors::filtering::base_filter &sample_filter = cuvs::neighbors::filtering::none_sample_filter{})#

void search(raft::resources const &res, cuvs::neighbors::cagra::search_params const &params, const cuvs::neighbors::cagra::index<half, uint32_t> &index, raft::device_matrix_view<const half, int64_t, raft::row_major> queries, raft::device_matrix_view<uint32_t, int64_t, raft::row_major> neighbors, raft::device_matrix_view<float, int64_t, raft::row_major> distances, const cuvs::neighbors::filtering::base_filter &sample_filter = cuvs::neighbors::filtering::none_sample_filter{})#

Search ANN using the constructed index.

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

Parameters:

res – [in] raft resources
params – [in] configure the search
index – [in] cagra index
queries – [in] a device matrix view to a row-major matrix [n_queries, index->dim()]
neighbors – [out] a device matrix view to the indices of the neighbors in the source dataset [n_queries, k]
distances – [out] a device matrix view to the distances to the selected neighbors [n_queries, k]
sample_filter – [in] an optional device filter function object that greenlights samples for a given query. (none_sample_filter for no filtering)

void search(raft::resources const &res, cuvs::neighbors::cagra::search_params const &params, const cuvs::neighbors::cagra::index<int8_t, uint32_t> &index, raft::device_matrix_view<const int8_t, int64_t, raft::row_major> queries, raft::device_matrix_view<uint32_t, int64_t, raft::row_major> neighbors, raft::device_matrix_view<float, int64_t, raft::row_major> distances, const cuvs::neighbors::filtering::base_filter &sample_filter = cuvs::neighbors::filtering::none_sample_filter{})#

Search ANN using the constructed index.

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

Parameters:

res – [in] raft resources
params – [in] configure the search
index – [in] cagra index
queries – [in] a device matrix view to a row-major matrix [n_queries, index->dim()]
neighbors – [out] a device matrix view to the indices of the neighbors in the source dataset [n_queries, k]
distances – [out] a device matrix view to the distances to the selected neighbors [n_queries, k]
sample_filter – [in] an optional device filter function object that greenlights samples for a given query. (none_sample_filter for no filtering)

void search(raft::resources const &res, cuvs::neighbors::cagra::search_params const &params, const cuvs::neighbors::cagra::index<uint8_t, uint32_t> &index, raft::device_matrix_view<const uint8_t, int64_t, raft::row_major> queries, raft::device_matrix_view<uint32_t, int64_t, raft::row_major> neighbors, raft::device_matrix_view<float, int64_t, raft::row_major> distances, const cuvs::neighbors::filtering::base_filter &sample_filter = cuvs::neighbors::filtering::none_sample_filter{})#

Search ANN using the constructed index.

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

Parameters:

res – [in] raft resources
params – [in] configure the search
index – [in] cagra index
queries – [in] a device matrix view to a row-major matrix [n_queries, index->dim()]
neighbors – [out] a device matrix view to the indices of the neighbors in the source dataset [n_queries, k]
distances – [out] a device matrix view to the distances to the selected neighbors [n_queries, k]
sample_filter – [in] an optional device filter function object that greenlights samples for a given query. (none_sample_filter for no filtering)

Index extend#

void extend(raft::resources const &handle, const cagra::extend_params &params, raft::device_matrix_view<const float, int64_t, raft::row_major> additional_dataset, cuvs::neighbors::cagra::index<float, uint32_t> &idx, std::optional<raft::device_matrix_view<float, int64_t, raft::layout_stride>> new_dataset_buffer_view = std::nullopt, std::optional<raft::device_matrix_view<uint32_t, int64_t>> new_graph_buffer_view = std::nullopt)#

Add new vectors to a CAGRA index.

Usage example:

using namespace raft::neighbors;
auto additional_dataset = raft::make_device_matrix<float, int64_t>(handle,add_size,dim);
// set_additional_dataset(additional_dataset.view());

cagra::extend_params params;
cagra::extend(res, params, raft::make_const_mdspan(additional_dataset.view()), index);

Parameters:

handle – [in] raft resources
params – [in] extend params
additional_dataset – [in] additional dataset on device memory
idx – [inout] CAGRA index
new_dataset_buffer_view – [out] memory buffer view for the dataset including the additional part. The data will be copied from the current index in this function. The num rows must be the sum of the original and additional datasets, cols must be the dimension of the dataset, and the stride must be the same as the original index dataset. This view will be stored in the output index. It is the caller’s responsibility to ensure that dataset stays alive as long as the index. This option is useful when users want to manage the memory space for the dataset themselves.
new_graph_buffer_view – [out] memory buffer view for the graph including the additional part. The data will be copied from the current index in this function. The num rows must be the sum of the original and additional datasets and cols must be the graph degree. This view will be stored in the output index. It is the caller’s responsibility to ensure that dataset stays alive as long as the index. This option is useful when users want to manage the memory space for the graph themselves.

void extend(raft::resources const &handle, const cagra::extend_params &params, raft::host_matrix_view<const float, int64_t, raft::row_major> additional_dataset, cuvs::neighbors::cagra::index<float, uint32_t> &idx, std::optional<raft::device_matrix_view<float, int64_t, raft::layout_stride>> new_dataset_buffer_view = std::nullopt, std::optional<raft::device_matrix_view<uint32_t, int64_t>> new_graph_buffer_view = std::nullopt)#

Add new vectors to a CAGRA index.

Usage example:

using namespace raft::neighbors;
auto additional_dataset = raft::make_host_matrix<float, int64_t>(handle,add_size,dim);
// set_additional_dataset(additional_dataset.view());

cagra::extend_params params;
cagra::extend(res, params, raft::make_const_mdspan(additional_dataset.view()), index);

Parameters:

handle – [in] raft resources
params – [in] extend params
additional_dataset – [in] additional dataset on host memory
idx – [inout] CAGRA index
new_dataset_buffer_view – [out] memory buffer view for the dataset including the additional part. The data will be copied from the current index in this function. The num rows must be the sum of the original and additional datasets, cols must be the dimension of the dataset, and the stride must be the same as the original index dataset. This view will be stored in the output index. It is the caller’s responsibility to ensure that dataset stays alive as long as the index. This option is useful when users want to manage the memory space for the dataset themselves.
new_graph_buffer_view – [out] memory buffer view for the graph including the additional part. The data will be copied from the current index in this function. The num rows must be the sum of the original and additional datasets and cols must be the graph degree. This view will be stored in the output index. It is the caller’s responsibility to ensure that dataset stays alive as long as the index. This option is useful when users want to manage the memory space for the graph themselves.

void extend(raft::resources const &handle, const cagra::extend_params &params, raft::device_matrix_view<const int8_t, int64_t, raft::row_major> additional_dataset, cuvs::neighbors::cagra::index<int8_t, uint32_t> &idx, std::optional<raft::device_matrix_view<int8_t, int64_t, raft::layout_stride>> new_dataset_buffer_view = std::nullopt, std::optional<raft::device_matrix_view<uint32_t, int64_t>> new_graph_buffer_view = std::nullopt)#

Add new vectors to a CAGRA index.

Usage example:

using namespace raft::neighbors;
auto additional_dataset = raft::make_device_matrix<int8_t, int64_t>(handle,add_size,dim);
// set_additional_dataset(additional_dataset.view());

cagra::extend_params params;
cagra::extend(res, params, raft::make_const_mdspan(additional_dataset.view()), index);

Parameters:

handle – [in] raft resources
params – [in] extend params
additional_dataset – [in] additional dataset on device memory
idx – [inout] CAGRA index
new_dataset_buffer_view – [out] memory buffer view for the dataset including the additional part. The data will be copied from the current index in this function. The num rows must be the sum of the original and additional datasets, cols must be the dimension of the dataset, and the stride must be the same as the original index dataset. This view will be stored in the output index. It is the caller’s responsibility to ensure that dataset stays alive as long as the index. This option is useful when users want to manage the memory space for the dataset themselves.
new_graph_buffer_view – [out] memory buffer view for the graph including the additional part. The data will be copied from the current index in this function. The num rows must be the sum of the original and additional datasets and cols must be the graph degree. This view will be stored in the output index. It is the caller’s responsibility to ensure that dataset stays alive as long as the index. This option is useful when users want to manage the memory space for the graph themselves.

void extend(raft::resources const &handle, const cagra::extend_params &params, raft::host_matrix_view<const int8_t, int64_t, raft::row_major> additional_dataset, cuvs::neighbors::cagra::index<int8_t, uint32_t> &idx, std::optional<raft::device_matrix_view<int8_t, int64_t, raft::layout_stride>> new_dataset_buffer_view = std::nullopt, std::optional<raft::device_matrix_view<uint32_t, int64_t>> new_graph_buffer_view = std::nullopt)#

Add new vectors to a CAGRA index.

Usage example:

using namespace raft::neighbors;
auto additional_dataset = raft::make_host_matrix<int8_t, int64_t>(handle,add_size,dim);
// set_additional_dataset(additional_dataset.view());

cagra::extend_params params;
cagra::extend(res, params, raft::make_const_mdspan(additional_dataset.view()), index);

Parameters:

handle – [in] raft resources
params – [in] extend params
additional_dataset – [in] additional dataset on host memory
idx – [inout] CAGRA index
new_dataset_buffer_view – [out] memory buffer view for the dataset including the additional part. The data will be copied from the current index in this function. The num rows must be the sum of the original and additional datasets, cols must be the dimension of the dataset, and the stride must be the same as the original index dataset. This view will be stored in the output index. It is the caller’s responsibility to ensure that dataset stays alive as long as the index. This option is useful when users want to manage the memory space for the dataset themselves.
new_graph_buffer_view – [out] memory buffer view for the graph including the additional part. The data will be copied from the current index in this function. The num rows must be the sum of the original and additional datasets and cols must be the graph degree. This view will be stored in the output index. It is the caller’s responsibility to ensure that dataset stays alive as long as the index. This option is useful when users want to manage the memory space for the graph themselves.

void extend(raft::resources const &handle, const cagra::extend_params &params, raft::device_matrix_view<const uint8_t, int64_t, raft::row_major> additional_dataset, cuvs::neighbors::cagra::index<uint8_t, uint32_t> &idx, std::optional<raft::device_matrix_view<uint8_t, int64_t, raft::layout_stride>> new_dataset_buffer_view = std::nullopt, std::optional<raft::device_matrix_view<uint32_t, int64_t>> new_graph_buffer_view = std::nullopt)#

Add new vectors to a CAGRA index.

Usage example:

using namespace raft::neighbors;
auto additional_dataset = raft::make_host_matrix<uint8_t, int64_t>(handle,add_size,dim);
// set_additional_dataset(additional_dataset.view());

cagra::extend_params params;
cagra::extend(res, params, raft::make_const_mdspan(additional_dataset.view()), index);

Parameters:

handle – [in] raft resources
params – [in] extend params
additional_dataset – [in] additional dataset on host memory
idx – [inout] CAGRA index
new_dataset_buffer_view – [out] memory buffer view for the dataset including the additional part. The data will be copied from the current index in this function. The num rows must be the sum of the original and additional datasets, cols must be the dimension of the dataset, and the stride must be the same as the original index dataset. This view will be stored in the output index. It is the caller’s responsibility to ensure that dataset stays alive as long as the index. This option is useful when users want to manage the memory space for the dataset themselves.
new_graph_buffer_view – [out] memory buffer view for the graph including the additional part. The data will be copied from the current index in this function. The num rows must be the sum of the original and additional datasets and cols must be the graph degree. This view will be stored in the output index. It is the caller’s responsibility to ensure that dataset stays alive as long as the index. This option is useful when users want to manage the memory space for the graph themselves.

void extend(raft::resources const &handle, const cagra::extend_params &params, raft::host_matrix_view<const uint8_t, int64_t, raft::row_major> additional_dataset, cuvs::neighbors::cagra::index<uint8_t, uint32_t> &idx, std::optional<raft::device_matrix_view<uint8_t, int64_t, raft::layout_stride>> new_dataset_buffer_view = std::nullopt, std::optional<raft::device_matrix_view<uint32_t, int64_t>> new_graph_buffer_view = std::nullopt)#

Add new vectors to a CAGRA index.

Usage example:

using namespace raft::neighbors;
auto additional_dataset = raft::make_host_matrix<uint8_t, int64_t>(handle,add_size,dim);
// set_additional_dataset(additional_dataset.view());

cagra::extend_params params;
cagra::extend(res, params, raft::make_const_mdspan(additional_dataset.view()), index);

Parameters:

handle – [in] raft resources
params – [in] extend params
additional_dataset – [in] additional dataset on host memory
idx – [inout] CAGRA index
new_dataset_buffer_view – [out] memory buffer view for the dataset including the additional part. The data will be copied from the current index in this function. The num rows must be the sum of the original and additional datasets, cols must be the dimension of the dataset, and the stride must be the same as the original index dataset. This view will be stored in the output index. It is the caller’s responsibility to ensure that dataset stays alive as long as the index. This option is useful when users want to manage the memory space for the dataset themselves.
new_graph_buffer_view – [out] memory buffer view for the graph including the additional part. The data will be copied from the current index in this function. The num rows must be the sum of the original and additional datasets and cols must be the graph degree. This view will be stored in the output index. It is the caller’s responsibility to ensure that dataset stays alive as long as the index. This option is useful when users want to manage the memory space for the graph themselves.

Index serialize#

void serialize(raft::resources const &handle, const std::string &filename, const cuvs::neighbors::cagra::index<float, uint32_t> &index, bool include_dataset = true)#

Save the index to file.