1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
|
#include "../../hnswlib/hnswlib.h"
int main() {
int dim = 16; // Dimension of the elements
int max_elements = 10000; // Maximum number of elements, should be known beforehand
int M = 16; // Tightly connected with internal dimensionality of the data
// strongly affects the memory consumption
int ef_construction = 200; // Controls index search speed/build speed tradeoff
// Initing index
hnswlib::L2Space space(dim);
hnswlib::HierarchicalNSW<float>* alg_hnsw = new hnswlib::HierarchicalNSW<float>(&space, max_elements, M, ef_construction);
// Generate random data
std::mt19937 rng;
rng.seed(47);
std::uniform_real_distribution<> distrib_real;
float* data = new float[dim * max_elements];
for (int i = 0; i < dim * max_elements; i++) {
data[i] = distrib_real(rng);
}
// Add data to index
for (int i = 0; i < max_elements; i++) {
alg_hnsw->addPoint(data + i * dim, i);
}
// Query the elements for themselves and measure recall
float correct = 0;
for (int i = 0; i < max_elements; i++) {
std::priority_queue<std::pair<float, hnswlib::labeltype>> result = alg_hnsw->searchKnn(data + i * dim, 1);
hnswlib::labeltype label = result.top().second;
if (label == i) correct++;
}
float recall = correct / max_elements;
std::cout << "Recall: " << recall << "\n";
// Serialize index
std::string hnsw_path = "hnsw.bin";
alg_hnsw->saveIndex(hnsw_path);
delete alg_hnsw;
// Deserialize index and check recall
alg_hnsw = new hnswlib::HierarchicalNSW<float>(&space, hnsw_path);
correct = 0;
for (int i = 0; i < max_elements; i++) {
std::priority_queue<std::pair<float, hnswlib::labeltype>> result = alg_hnsw->searchKnn(data + i * dim, 1);
hnswlib::labeltype label = result.top().second;
if (label == i) correct++;
}
recall = (float)correct / max_elements;
std::cout << "Recall of deserialized index: " << recall << "\n";
delete[] data;
delete alg_hnsw;
return 0;
}
|
