Forward-merge branch-24.10 into branch-24.12

cpp/CMakeLists.txt

@@ -202,6 +202,7 @@ endif()
 add_library(
   cuvs-cagra-search STATIC
   src/neighbors/cagra_search_float.cu
+  src/neighbors/cagra_search_half.cu
   src/neighbors/cagra_search_int8.cu
   src/neighbors/cagra_search_uint8.cu
   src/neighbors/detail/cagra/compute_distance.cu
@@ -257,14 +258,10 @@ add_library(
   src/neighbors/detail/cagra/search_multi_cta_half_uint32.cu
   src/neighbors/detail/cagra/search_multi_cta_int8_uint32.cu
   src/neighbors/detail/cagra/search_multi_cta_uint8_uint32.cu
-  src/neighbors/detail/cagra/search_multi_cta_float_uint64.cu
-  src/neighbors/detail/cagra/search_multi_cta_half_uint64.cu
   src/neighbors/detail/cagra/search_single_cta_float_uint32.cu
   src/neighbors/detail/cagra/search_single_cta_half_uint32.cu
   src/neighbors/detail/cagra/search_single_cta_int8_uint32.cu
   src/neighbors/detail/cagra/search_single_cta_uint8_uint32.cu
-  src/neighbors/detail/cagra/search_single_cta_float_uint64.cu
-  src/neighbors/detail/cagra/search_single_cta_half_uint64.cu
 )
 
 file(GLOB_RECURSE compute_distance_sources "src/neighbors/detail/cagra/compute_distance_*.cu")
@@ -350,13 +347,15 @@ add_library(
   src/distance/pairwise_distance.cu
   src/neighbors/brute_force.cu
   src/neighbors/cagra_build_float.cu
+  src/neighbors/cagra_build_half.cu
   src/neighbors/cagra_build_int8.cu
   src/neighbors/cagra_build_uint8.cu
   src/neighbors/cagra_extend_float.cu
   src/neighbors/cagra_extend_int8.cu
   src/neighbors/cagra_extend_uint8.cu
   src/neighbors/cagra_optimize.cu
   src/neighbors/cagra_serialize_float.cu
+  src/neighbors/cagra_serialize_half.cu
   src/neighbors/cagra_serialize_int8.cu
   src/neighbors/cagra_serialize_uint8.cu
   src/neighbors/detail/cagra/cagra_build.cpp
@@ -378,6 +377,7 @@ add_library(
   src/neighbors/ivf_pq/ivf_pq_serialize.cu
   src/neighbors/ivf_pq/ivf_pq_deserialize.cu
   src/neighbors/ivf_pq/detail/ivf_pq_build_extend_float_int64_t.cu
+  src/neighbors/ivf_pq/detail/ivf_pq_build_extend_half_int64_t.cu
   src/neighbors/ivf_pq/detail/ivf_pq_build_extend_int8_t_int64_t.cu
   src/neighbors/ivf_pq/detail/ivf_pq_build_extend_uint8_t_int64_t.cu
   src/neighbors/ivf_pq/detail/ivf_pq_compute_similarity_half_fp8_false.cu
@@ -395,13 +395,16 @@ add_library(
   src/neighbors/ivf_pq/detail/ivf_pq_compute_similarity_float_fp8_false_bitset64.cu
   src/neighbors/ivf_pq/detail/ivf_pq_compute_similarity_float_fp8_true_bitset64.cu
   src/neighbors/ivf_pq/detail/ivf_pq_search_float_int64_t.cu
+  src/neighbors/ivf_pq/detail/ivf_pq_search_half_int64_t.cu
   src/neighbors/ivf_pq/detail/ivf_pq_search_int8_t_int64_t.cu
   src/neighbors/ivf_pq/detail/ivf_pq_search_uint8_t_int64_t.cu
   src/neighbors/ivf_pq/detail/ivf_pq_search_with_filter_float_int64_t.cu
+  src/neighbors/ivf_pq/detail/ivf_pq_search_with_filter_half_int64_t.cu
   src/neighbors/ivf_pq/detail/ivf_pq_search_with_filter_int8_t_int64_t.cu
   src/neighbors/ivf_pq/detail/ivf_pq_search_with_filter_uint8_t_int64_t.cu
   src/neighbors/nn_descent.cu
   src/neighbors/nn_descent_float.cu
+  src/neighbors/nn_descent_half.cu
   src/neighbors/nn_descent_int8.cu
   src/neighbors/nn_descent_uint8.cu
   src/neighbors/refine/detail/refine_device_float_float.cu

cpp/bench/ann/src/common/benchmark.hpp

@@ -687,17 +687,17 @@ inline auto run_main(int argc, char** argv) -> int
                               override_kv,
                               metric_objective,
                               threads);
-    // } else if (dtype == "half") {
-    //   dispatch_benchmark<half>(cmdline
-    //                            conf,
-    //                            force_overwrite,
-    //                            build_mode,
-    //                            search_mode,
-    //                            data_prefix,
-    //                            index_prefix,
-    //                            override_kv,
-    //                            metric_objective,
-    //                            threads);
+  } else if (dtype == "half") {
+    dispatch_benchmark<half>(cmdline,
+                             conf,
+                             force_overwrite,
+                             build_mode,
+                             search_mode,
+                             data_prefix,
+                             index_prefix,
+                             override_kv,
+                             metric_objective,
+                             threads);
   } else if (dtype == "uint8") {
     dispatch_benchmark<std::uint8_t>(cmdline,
                                      conf,

cpp/bench/ann/src/cuvs/cuvs_benchmark.cu

@@ -121,7 +121,7 @@ auto create_search_param(const std::string& algo_name, const nlohmann::json& con
 };  // namespace cuvs::bench
 
 REGISTER_ALGO_INSTANCE(float);
-// REGISTER_ALGO_INSTANCE(half);
+REGISTER_ALGO_INSTANCE(half);
 REGISTER_ALGO_INSTANCE(std::int8_t);
 REGISTER_ALGO_INSTANCE(std::uint8_t);
 

cpp/bench/ann/src/cuvs/cuvs_cagra_half.cu

@@ -16,5 +16,5 @@
 #include "cuvs_cagra_wrapper.h"
 
 namespace cuvs::bench {
-// template class cuvs_cagra<half, uint32_t>;
+template class cuvs_cagra<half, uint32_t>;
 }  // namespace cuvs::bench

cpp/bench/ann/src/cuvs/cuvs_cagra_wrapper.h

@@ -289,7 +289,11 @@ void cuvs_cagra<T, IdxT>::save(const std::string& file) const
 template <typename T, typename IdxT>
 void cuvs_cagra<T, IdxT>::save_to_hnswlib(const std::string& file) const
 {
-  cuvs::neighbors::cagra::serialize_to_hnswlib(handle_, file, *index_);
+  if constexpr (!std::is_same_v<T, half>) {
+    cuvs::neighbors::cagra::serialize_to_hnswlib(handle_, file, *index_);
+  } else {
+    RAFT_FAIL("Cannot save fp16 index to hnswlib format");
+  }
 }
 
 template <typename T, typename IdxT>

cpp/bench/ann/src/cuvs/cuvs_ivf_pq.cu

@@ -17,7 +17,7 @@
 
 namespace cuvs::bench {
 template class cuvs_ivf_pq<float, int64_t>;
-// template class cuvs_ivf_pq<half, int64_t>;
+template class cuvs_ivf_pq<half, int64_t>;
 template class cuvs_ivf_pq<uint8_t, int64_t>;
 template class cuvs_ivf_pq<int8_t, int64_t>;
 }  // namespace cuvs::bench

cpp/include/cuvs/neighbors/cagra.hpp

@@ -613,6 +613,78 @@ auto build(raft::resources const& res,
  *
  * @return 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>;
+
+/**
+ * @brief 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
+ *
+ * Usage example:
+ * @code{.cpp}
+ *   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);
+ * @endcode
+ *
+ * @param[in] res
+ * @param[in] params parameters for building the index
+ * @param[in] dataset a matrix view (host) to a row-major matrix [n_rows, dim]
+ *
+ * @return 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>;
+
+/**
+ * @brief 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
+ *
+ * Usage example:
+ * @code{.cpp}
+ *   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);
+ * @endcode
+ *
+ * @param[in] res
+ * @param[in] params parameters for building the index
+ * @param[in] dataset a matrix view (device) to a row-major matrix [n_rows, dim]
+ *
+ * @return 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)
@@ -975,9 +1047,6 @@ void extend(
  *
  * See the [cagra::build](#cagra::build) documentation for a usage example.
  *
- * @tparam T data element type
- * @tparam IdxT type of the indices
- *
  * @param[in] res raft resources
  * @param[in] params configure the search
  * @param[in] idx cagra index
@@ -1000,8 +1069,26 @@ void search(raft::resources const& res,
  *
  * See the [cagra::build](#cagra::build) documentation for a usage example.
  *
- * @tparam T data element type
- * @tparam IdxT type of the indices
+ * @param[in] res raft resources
+ * @param[in] params configure the search
+ * @param[in] index cagra index
+ * @param[in] queries a device matrix view to a row-major matrix [n_queries, index->dim()]
+ * @param[out] neighbors a device matrix view to the indices of the neighbors in the source dataset
+ * [n_queries, k]
+ * @param[out] distances a device matrix view to the distances to the selected neighbors [n_queries,
+ * k]
+ */
+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);
+
+/**
+ * @brief Search ANN using the constructed index.
+ *
+ * See the [cagra::build](#cagra::build) documentation for a usage example.
  *
  * @param[in] res raft resources
  * @param[in] params configure the search
@@ -1024,9 +1111,6 @@ void search(raft::resources const& res,
  *
  * See the [cagra::build](#cagra::build) documentation for a usage example.
  *
- * @tparam T data element type
- * @tparam IdxT type of the indices
- *
  * @param[in] res raft resources
  * @param[in] params configure the search
  * @param[in] index cagra index
@@ -1156,6 +1240,111 @@ void serialize(raft::resources const& handle,
 void deserialize(raft::resources const& handle,
                  std::istream& is,
                  cuvs::neighbors::cagra::index<float, uint32_t>* index);
+/**
+ * Save the index to file.
+ *
+ * Experimental, both the API and the serialization format are subject to change.
+ *
+ * @code{.cpp}
+ * #include <raft/core/resources.hpp>
+ * #include <cuvs/neighbors/cagra.hpp>
+ *
+ * raft::resources handle;
+ *
+ * // create a string with a filepath
+ * std::string filename("/path/to/index");
+ * // create an index with `auto index = cuvs::neighbors::cagra::build(...);`
+ * cuvs::neighbors::cagra::serialize(handle, filename, index);
+ * @endcode
+ *
+ * @param[in] handle the raft handle
+ * @param[in] filename the file name for saving the index
+ * @param[in] index CAGRA index
+ * @param[in] include_dataset Whether or not to write out the dataset to the file.
+ *
+ */
+void serialize(raft::resources const& handle,
+               const std::string& filename,
+               const cuvs::neighbors::cagra::index<half, uint32_t>& index,
+               bool include_dataset = true);
+
+/**
+ * Load index from file.
+ *
+ * Experimental, both the API and the serialization format are subject to change.
+ *
+ * @code{.cpp}
+ * #include <raft/core/resources.hpp>
+ * #include <cuvs/neighbors/cagra.hpp>
+ *
+ * raft::resources handle;
+ *
+ * // create a string with a filepath
+ * std::string filename("/path/to/index");
+
+ * cuvs::neighbors::cagra::index<half, uint32_t> index;
+ * cuvs::neighbors::cagra::deserialize(handle, filename, &index);
+ * @endcode
+ *
+ * @param[in] handle the raft handle
+ * @param[in] filename the name of the file that stores the index
+ * @param[out] index the cagra index
+ */
+void deserialize(raft::resources const& handle,
+                 const std::string& filename,
+                 cuvs::neighbors::cagra::index<half, uint32_t>* index);
+
+/**
+ * Write the index to an output stream
+ *
+ * Experimental, both the API and the serialization format are subject to change.
+ *
+ * @code{.cpp}
+ * #include <raft/core/resources.hpp>
+ * #include <cuvs/neighbors/cagra.hpp>
+ *
+ * raft::resources handle;
+ *
+ * // create an output stream
+ * std::ostream os(std::cout.rdbuf());
+ * // create an index with `auto index = cuvs::neighbors::cagra::build(...);`
+ * cuvs::neighbors::cagra::serialize(handle, os, index);
+ * @endcode
+ *
+ * @param[in] handle the raft handle
+ * @param[in] os output stream
+ * @param[in] index CAGRA index
+ * @param[in] include_dataset Whether or not to write out the dataset to the file.
+ */
+void serialize(raft::resources const& handle,
+               std::ostream& os,
+               const cuvs::neighbors::cagra::index<half, uint32_t>& index,
+               bool include_dataset = true);
+
+/**
+ * Load index from input stream
+ *
+ * Experimental, both the API and the serialization format are subject to change.
+ *
+ * @code{.cpp}
+ * #include <raft/core/resources.hpp>
+ * #include <cuvs/neighbors/cagra.hpp>
+ *
+ * raft::resources handle;
+ *
+ * // create an input stream
+ * std::istream is(std::cin.rdbuf());
+ * cuvs::neighbors::cagra::index<half, uint32_t> index;
+ * cuvs::neighbors::cagra::deserialize(handle, is, &index);
+ * @endcode
+ *
+ * @param[in] handle the raft handle
+ * @param[in] is input stream
+ * @param[out] index the cagra index
+ */
+void deserialize(raft::resources const& handle,
+                 std::istream& is,
+                 cuvs::neighbors::cagra::index<half, uint32_t>* index);
 
 /**
  * Save the index to file.