Fix ef setting in HNSW wrapper (#2367)

Closes #2363 

Bugs fixed: 

1. Setting `ef` in search, it was not being set at all before
2. `from_cagra` used a hard coded filename to serialize CAGRA graph and deserialize to HNSW graph. The PR changes the hardcoded filename to a random string so that multiple graphs may be converted concurrently

cc @Presburger thank you for reporting these bugs

Authors:
  - Divye Gala (https://github.com/divyegala)

Approvers:
  - Corey J. Nolet (https://github.com/cjnolet)

URL: #2367

cpp/include/raft/neighbors/detail/hnsw.hpp

@@ -53,6 +53,7 @@ void search(raft::resources const& res,
             raft::host_matrix_view<uint64_t, int64_t, row_major> neighbors,
             raft::host_matrix_view<float, int64_t, row_major> distances)
 {
+  idx.set_ef(params.ef);
   auto const* hnswlib_index =
     reinterpret_cast<hnswlib::HierarchicalNSW<typename hnsw_dist_t<T>::type> const*>(
       idx.get_index());

cpp/include/raft/neighbors/detail/hnsw_types.hpp

@@ -93,6 +93,11 @@ struct index_impl : index<T> {
   */
   auto get_index() const -> void const* override { return appr_alg_.get(); }
 
+  /**
+  @brief Set ef for search
+  */
+  void set_ef(int ef) const override { appr_alg_->ef_ = ef; }
+
  private:
   std::unique_ptr<hnswlib::HierarchicalNSW<typename hnsw_dist_t<T>::type>> appr_alg_;
   std::unique_ptr<hnswlib::SpaceInterface<typename hnsw_dist_t<T>::type>> space_;

cpp/include/raft/neighbors/hnsw.hpp

@@ -35,7 +35,7 @@ namespace raft::neighbors::hnsw {
 
 /**
  * @brief Construct an hnswlib base-layer-only index from a CAGRA index
- * NOTE: 1. This method uses the filesystem to write the CAGRA index in `/tmp/cagra_index.bin`
+ * NOTE: 1. This method uses the filesystem to write the CAGRA index in `/tmp/<random_number>.bin`
  * before reading it as an hnswlib index, then deleting the temporary file.
  *       2. This function is only offered as a compiled symbol in `libraft.so`
  *

cpp/include/raft/neighbors/hnsw_types.hpp

@@ -62,6 +62,11 @@ struct index : ann::index {
 
   auto metric() const -> raft::distance::DistanceType { return metric_; }
 
+  /**
+  @brief Set ef for search
+  */
+  virtual void set_ef(int ef) const;
+
  private:
   int dim_;
   raft::distance::DistanceType metric_;

cpp/src/raft_runtime/neighbors/hnsw.cpp

@@ -21,14 +21,20 @@
 #include <raft_runtime/neighbors/hnsw.hpp>
 
 #include <filesystem>
+#include <random>
+#include <string>
 
 namespace raft::neighbors::hnsw {
 #define RAFT_INST_HNSW(T)                                                               \
   template <>                                                                           \
   std::unique_ptr<raft::neighbors::hnsw::index<T>> from_cagra(                          \
     raft::resources const& res, raft::neighbors::cagra::index<T, uint32_t> cagra_index) \
   {                                                                                     \
-    std::string filepath = "/tmp/cagra_index.bin";                                      \
+    std::random_device dev;                                                             \
+    std::mt19937 rng(dev());                                                            \
+    std::uniform_int_distribution<std::mt19937::result_type> dist(0);                   \
+    auto uuid            = std::to_string(dist(rng));                                   \
+    std::string filepath = "/tmp/" + uuid + ".bin";                                     \
     raft::runtime::neighbors::cagra::serialize_to_hnswlib(res, filepath, cagra_index);  \
     auto hnsw_index = raft::runtime::neighbors::hnsw::deserialize_file<T>(              \
       res, filepath, cagra_index.dim(), cagra_index.metric());                          \

python/pylibraft/pylibraft/neighbors/hnsw.pyx

@@ -52,6 +52,7 @@ from pylibraft.common.mdspan cimport (
 from pylibraft.neighbors.common cimport _get_metric_string
 
 import os
+import uuid
 
 import numpy as np
 
@@ -292,7 +293,7 @@ def from_cagra(Index index, handle=None):
     Returns an hnswlib base-layer-only index from a CAGRA index.
 
     NOTE: This method uses the filesystem to write the CAGRA index in
-          `/tmp/cagra_index.bin` before reading it as an hnswlib index,
+          `/tmp/<random_number>.bin` before reading it as an hnswlib index,
           then deleting the temporary file.
 
     Saving / loading the index is experimental. The serialization format is
@@ -320,7 +321,8 @@ def from_cagra(Index index, handle=None):
     >>> # Serialize the CAGRA index to hnswlib base layer only index format
     >>> hnsw_index = hnsw.from_cagra(index, handle=handle)
     """
-    filename = "/tmp/cagra_index.bin"
+    uuid_num = uuid.uuid4()
+    filename = f"/tmp/{uuid_num}.bin"
     save(filename, index, handle=handle)
     hnsw_index = load(filename, index.dim, np.dtype(index.active_index_type),
                       _get_metric_string(index.metric), handle=handle)