faiss 0.2.0 → 0.2.4
Sign up to get free protection for your applications and to get access to all the features.
- checksums.yaml +4 -4
- data/CHANGELOG.md +16 -0
- data/LICENSE.txt +1 -1
- data/README.md +7 -7
- data/ext/faiss/extconf.rb +6 -3
- data/ext/faiss/numo.hpp +4 -4
- data/ext/faiss/utils.cpp +1 -1
- data/ext/faiss/utils.h +1 -1
- data/lib/faiss/version.rb +1 -1
- data/vendor/faiss/faiss/AutoTune.cpp +292 -291
- data/vendor/faiss/faiss/AutoTune.h +55 -56
- data/vendor/faiss/faiss/Clustering.cpp +365 -194
- data/vendor/faiss/faiss/Clustering.h +102 -35
- data/vendor/faiss/faiss/IVFlib.cpp +171 -195
- data/vendor/faiss/faiss/IVFlib.h +48 -51
- data/vendor/faiss/faiss/Index.cpp +85 -103
- data/vendor/faiss/faiss/Index.h +54 -48
- data/vendor/faiss/faiss/Index2Layer.cpp +126 -224
- data/vendor/faiss/faiss/Index2Layer.h +22 -36
- data/vendor/faiss/faiss/IndexAdditiveQuantizer.cpp +407 -0
- data/vendor/faiss/faiss/IndexAdditiveQuantizer.h +195 -0
- data/vendor/faiss/faiss/IndexBinary.cpp +45 -37
- data/vendor/faiss/faiss/IndexBinary.h +140 -132
- data/vendor/faiss/faiss/IndexBinaryFlat.cpp +73 -53
- data/vendor/faiss/faiss/IndexBinaryFlat.h +29 -24
- data/vendor/faiss/faiss/IndexBinaryFromFloat.cpp +46 -43
- data/vendor/faiss/faiss/IndexBinaryFromFloat.h +16 -15
- data/vendor/faiss/faiss/IndexBinaryHNSW.cpp +215 -232
- data/vendor/faiss/faiss/IndexBinaryHNSW.h +25 -24
- data/vendor/faiss/faiss/IndexBinaryHash.cpp +182 -177
- data/vendor/faiss/faiss/IndexBinaryHash.h +41 -34
- data/vendor/faiss/faiss/IndexBinaryIVF.cpp +489 -461
- data/vendor/faiss/faiss/IndexBinaryIVF.h +97 -68
- data/vendor/faiss/faiss/IndexFlat.cpp +115 -176
- data/vendor/faiss/faiss/IndexFlat.h +42 -59
- data/vendor/faiss/faiss/IndexFlatCodes.cpp +67 -0
- data/vendor/faiss/faiss/IndexFlatCodes.h +47 -0
- data/vendor/faiss/faiss/IndexHNSW.cpp +372 -348
- data/vendor/faiss/faiss/IndexHNSW.h +57 -41
- data/vendor/faiss/faiss/IndexIVF.cpp +545 -453
- data/vendor/faiss/faiss/IndexIVF.h +169 -118
- data/vendor/faiss/faiss/IndexIVFAdditiveQuantizer.cpp +316 -0
- data/vendor/faiss/faiss/IndexIVFAdditiveQuantizer.h +121 -0
- data/vendor/faiss/faiss/IndexIVFFlat.cpp +247 -252
- data/vendor/faiss/faiss/IndexIVFFlat.h +48 -51
- data/vendor/faiss/faiss/IndexIVFPQ.cpp +459 -517
- data/vendor/faiss/faiss/IndexIVFPQ.h +75 -67
- data/vendor/faiss/faiss/IndexIVFPQFastScan.cpp +406 -372
- data/vendor/faiss/faiss/IndexIVFPQFastScan.h +82 -57
- data/vendor/faiss/faiss/IndexIVFPQR.cpp +104 -102
- data/vendor/faiss/faiss/IndexIVFPQR.h +33 -28
- data/vendor/faiss/faiss/IndexIVFSpectralHash.cpp +163 -150
- data/vendor/faiss/faiss/IndexIVFSpectralHash.h +38 -25
- data/vendor/faiss/faiss/IndexLSH.cpp +66 -113
- data/vendor/faiss/faiss/IndexLSH.h +20 -38
- data/vendor/faiss/faiss/IndexLattice.cpp +42 -56
- data/vendor/faiss/faiss/IndexLattice.h +11 -16
- data/vendor/faiss/faiss/IndexNNDescent.cpp +229 -0
- data/vendor/faiss/faiss/IndexNNDescent.h +72 -0
- data/vendor/faiss/faiss/IndexNSG.cpp +301 -0
- data/vendor/faiss/faiss/IndexNSG.h +85 -0
- data/vendor/faiss/faiss/IndexPQ.cpp +387 -495
- data/vendor/faiss/faiss/IndexPQ.h +64 -82
- data/vendor/faiss/faiss/IndexPQFastScan.cpp +143 -170
- data/vendor/faiss/faiss/IndexPQFastScan.h +46 -32
- data/vendor/faiss/faiss/IndexPreTransform.cpp +120 -150
- data/vendor/faiss/faiss/IndexPreTransform.h +33 -36
- data/vendor/faiss/faiss/IndexRefine.cpp +139 -127
- data/vendor/faiss/faiss/IndexRefine.h +32 -23
- data/vendor/faiss/faiss/IndexReplicas.cpp +147 -153
- data/vendor/faiss/faiss/IndexReplicas.h +62 -56
- data/vendor/faiss/faiss/IndexScalarQuantizer.cpp +111 -172
- data/vendor/faiss/faiss/IndexScalarQuantizer.h +41 -59
- data/vendor/faiss/faiss/IndexShards.cpp +256 -240
- data/vendor/faiss/faiss/IndexShards.h +85 -73
- data/vendor/faiss/faiss/MatrixStats.cpp +112 -97
- data/vendor/faiss/faiss/MatrixStats.h +7 -10
- data/vendor/faiss/faiss/MetaIndexes.cpp +135 -157
- data/vendor/faiss/faiss/MetaIndexes.h +40 -34
- data/vendor/faiss/faiss/MetricType.h +7 -7
- data/vendor/faiss/faiss/VectorTransform.cpp +654 -475
- data/vendor/faiss/faiss/VectorTransform.h +64 -89
- data/vendor/faiss/faiss/clone_index.cpp +78 -73
- data/vendor/faiss/faiss/clone_index.h +4 -9
- data/vendor/faiss/faiss/gpu/GpuAutoTune.cpp +33 -38
- data/vendor/faiss/faiss/gpu/GpuAutoTune.h +11 -9
- data/vendor/faiss/faiss/gpu/GpuCloner.cpp +198 -171
- data/vendor/faiss/faiss/gpu/GpuCloner.h +53 -35
- data/vendor/faiss/faiss/gpu/GpuClonerOptions.cpp +12 -14
- data/vendor/faiss/faiss/gpu/GpuClonerOptions.h +27 -25
- data/vendor/faiss/faiss/gpu/GpuDistance.h +116 -112
- data/vendor/faiss/faiss/gpu/GpuFaissAssert.h +1 -2
- data/vendor/faiss/faiss/gpu/GpuIcmEncoder.h +60 -0
- data/vendor/faiss/faiss/gpu/GpuIndex.h +134 -137
- data/vendor/faiss/faiss/gpu/GpuIndexBinaryFlat.h +76 -73
- data/vendor/faiss/faiss/gpu/GpuIndexFlat.h +173 -162
- data/vendor/faiss/faiss/gpu/GpuIndexIVF.h +67 -64
- data/vendor/faiss/faiss/gpu/GpuIndexIVFFlat.h +89 -86
- data/vendor/faiss/faiss/gpu/GpuIndexIVFPQ.h +150 -141
- data/vendor/faiss/faiss/gpu/GpuIndexIVFScalarQuantizer.h +101 -103
- data/vendor/faiss/faiss/gpu/GpuIndicesOptions.h +17 -16
- data/vendor/faiss/faiss/gpu/GpuResources.cpp +116 -128
- data/vendor/faiss/faiss/gpu/GpuResources.h +182 -186
- data/vendor/faiss/faiss/gpu/StandardGpuResources.cpp +433 -422
- data/vendor/faiss/faiss/gpu/StandardGpuResources.h +131 -130
- data/vendor/faiss/faiss/gpu/impl/InterleavedCodes.cpp +468 -456
- data/vendor/faiss/faiss/gpu/impl/InterleavedCodes.h +25 -19
- data/vendor/faiss/faiss/gpu/impl/RemapIndices.cpp +22 -20
- data/vendor/faiss/faiss/gpu/impl/RemapIndices.h +9 -8
- data/vendor/faiss/faiss/gpu/perf/IndexWrapper-inl.h +39 -44
- data/vendor/faiss/faiss/gpu/perf/IndexWrapper.h +16 -14
- data/vendor/faiss/faiss/gpu/perf/PerfClustering.cpp +77 -71
- data/vendor/faiss/faiss/gpu/perf/PerfIVFPQAdd.cpp +109 -88
- data/vendor/faiss/faiss/gpu/perf/WriteIndex.cpp +75 -64
- data/vendor/faiss/faiss/gpu/test/TestCodePacking.cpp +230 -215
- data/vendor/faiss/faiss/gpu/test/TestGpuIndexBinaryFlat.cpp +80 -86
- data/vendor/faiss/faiss/gpu/test/TestGpuIndexFlat.cpp +284 -277
- data/vendor/faiss/faiss/gpu/test/TestGpuIndexIVFFlat.cpp +416 -416
- data/vendor/faiss/faiss/gpu/test/TestGpuIndexIVFPQ.cpp +611 -517
- data/vendor/faiss/faiss/gpu/test/TestGpuIndexIVFScalarQuantizer.cpp +166 -164
- data/vendor/faiss/faiss/gpu/test/TestGpuMemoryException.cpp +61 -53
- data/vendor/faiss/faiss/gpu/test/TestUtils.cpp +274 -238
- data/vendor/faiss/faiss/gpu/test/TestUtils.h +73 -57
- data/vendor/faiss/faiss/gpu/test/demo_ivfpq_indexing_gpu.cpp +47 -50
- data/vendor/faiss/faiss/gpu/utils/DeviceUtils.h +79 -72
- data/vendor/faiss/faiss/gpu/utils/StackDeviceMemory.cpp +140 -146
- data/vendor/faiss/faiss/gpu/utils/StackDeviceMemory.h +69 -71
- data/vendor/faiss/faiss/gpu/utils/StaticUtils.h +21 -16
- data/vendor/faiss/faiss/gpu/utils/Timer.cpp +25 -29
- data/vendor/faiss/faiss/gpu/utils/Timer.h +30 -29
- data/vendor/faiss/faiss/impl/AdditiveQuantizer.cpp +503 -0
- data/vendor/faiss/faiss/impl/AdditiveQuantizer.h +175 -0
- data/vendor/faiss/faiss/impl/AuxIndexStructures.cpp +90 -120
- data/vendor/faiss/faiss/impl/AuxIndexStructures.h +81 -65
- data/vendor/faiss/faiss/impl/FaissAssert.h +73 -58
- data/vendor/faiss/faiss/impl/FaissException.cpp +56 -48
- data/vendor/faiss/faiss/impl/FaissException.h +41 -29
- data/vendor/faiss/faiss/impl/HNSW.cpp +606 -617
- data/vendor/faiss/faiss/impl/HNSW.h +179 -200
- data/vendor/faiss/faiss/impl/LocalSearchQuantizer.cpp +855 -0
- data/vendor/faiss/faiss/impl/LocalSearchQuantizer.h +244 -0
- data/vendor/faiss/faiss/impl/NNDescent.cpp +487 -0
- data/vendor/faiss/faiss/impl/NNDescent.h +154 -0
- data/vendor/faiss/faiss/impl/NSG.cpp +679 -0
- data/vendor/faiss/faiss/impl/NSG.h +199 -0
- data/vendor/faiss/faiss/impl/PolysemousTraining.cpp +484 -454
- data/vendor/faiss/faiss/impl/PolysemousTraining.h +52 -55
- data/vendor/faiss/faiss/impl/ProductQuantizer-inl.h +26 -47
- data/vendor/faiss/faiss/impl/ProductQuantizer.cpp +469 -459
- data/vendor/faiss/faiss/impl/ProductQuantizer.h +76 -87
- data/vendor/faiss/faiss/impl/ResidualQuantizer.cpp +758 -0
- data/vendor/faiss/faiss/impl/ResidualQuantizer.h +188 -0
- data/vendor/faiss/faiss/impl/ResultHandler.h +96 -132
- data/vendor/faiss/faiss/impl/ScalarQuantizer.cpp +647 -707
- data/vendor/faiss/faiss/impl/ScalarQuantizer.h +48 -46
- data/vendor/faiss/faiss/impl/ThreadedIndex-inl.h +129 -131
- data/vendor/faiss/faiss/impl/ThreadedIndex.h +61 -55
- data/vendor/faiss/faiss/impl/index_read.cpp +631 -480
- data/vendor/faiss/faiss/impl/index_write.cpp +547 -407
- data/vendor/faiss/faiss/impl/io.cpp +76 -95
- data/vendor/faiss/faiss/impl/io.h +31 -41
- data/vendor/faiss/faiss/impl/io_macros.h +60 -29
- data/vendor/faiss/faiss/impl/kmeans1d.cpp +301 -0
- data/vendor/faiss/faiss/impl/kmeans1d.h +48 -0
- data/vendor/faiss/faiss/impl/lattice_Zn.cpp +137 -186
- data/vendor/faiss/faiss/impl/lattice_Zn.h +40 -51
- data/vendor/faiss/faiss/impl/platform_macros.h +29 -8
- data/vendor/faiss/faiss/impl/pq4_fast_scan.cpp +77 -124
- data/vendor/faiss/faiss/impl/pq4_fast_scan.h +39 -48
- data/vendor/faiss/faiss/impl/pq4_fast_scan_search_1.cpp +41 -52
- data/vendor/faiss/faiss/impl/pq4_fast_scan_search_qbs.cpp +80 -117
- data/vendor/faiss/faiss/impl/simd_result_handlers.h +109 -137
- data/vendor/faiss/faiss/index_factory.cpp +619 -397
- data/vendor/faiss/faiss/index_factory.h +8 -6
- data/vendor/faiss/faiss/index_io.h +23 -26
- data/vendor/faiss/faiss/invlists/BlockInvertedLists.cpp +67 -75
- data/vendor/faiss/faiss/invlists/BlockInvertedLists.h +22 -24
- data/vendor/faiss/faiss/invlists/DirectMap.cpp +96 -112
- data/vendor/faiss/faiss/invlists/DirectMap.h +29 -33
- data/vendor/faiss/faiss/invlists/InvertedLists.cpp +307 -364
- data/vendor/faiss/faiss/invlists/InvertedLists.h +151 -151
- data/vendor/faiss/faiss/invlists/InvertedListsIOHook.cpp +29 -34
- data/vendor/faiss/faiss/invlists/InvertedListsIOHook.h +17 -18
- data/vendor/faiss/faiss/invlists/OnDiskInvertedLists.cpp +257 -293
- data/vendor/faiss/faiss/invlists/OnDiskInvertedLists.h +50 -45
- data/vendor/faiss/faiss/python/python_callbacks.cpp +23 -26
- data/vendor/faiss/faiss/python/python_callbacks.h +9 -16
- data/vendor/faiss/faiss/utils/AlignedTable.h +79 -44
- data/vendor/faiss/faiss/utils/Heap.cpp +40 -48
- data/vendor/faiss/faiss/utils/Heap.h +186 -209
- data/vendor/faiss/faiss/utils/WorkerThread.cpp +67 -76
- data/vendor/faiss/faiss/utils/WorkerThread.h +32 -33
- data/vendor/faiss/faiss/utils/distances.cpp +305 -312
- data/vendor/faiss/faiss/utils/distances.h +170 -122
- data/vendor/faiss/faiss/utils/distances_simd.cpp +498 -508
- data/vendor/faiss/faiss/utils/extra_distances-inl.h +117 -0
- data/vendor/faiss/faiss/utils/extra_distances.cpp +113 -232
- data/vendor/faiss/faiss/utils/extra_distances.h +30 -29
- data/vendor/faiss/faiss/utils/hamming-inl.h +260 -209
- data/vendor/faiss/faiss/utils/hamming.cpp +375 -469
- data/vendor/faiss/faiss/utils/hamming.h +62 -85
- data/vendor/faiss/faiss/utils/ordered_key_value.h +16 -18
- data/vendor/faiss/faiss/utils/partitioning.cpp +393 -318
- data/vendor/faiss/faiss/utils/partitioning.h +26 -21
- data/vendor/faiss/faiss/utils/quantize_lut.cpp +78 -66
- data/vendor/faiss/faiss/utils/quantize_lut.h +22 -20
- data/vendor/faiss/faiss/utils/random.cpp +39 -63
- data/vendor/faiss/faiss/utils/random.h +13 -16
- data/vendor/faiss/faiss/utils/simdlib.h +4 -2
- data/vendor/faiss/faiss/utils/simdlib_avx2.h +88 -85
- data/vendor/faiss/faiss/utils/simdlib_emulated.h +226 -165
- data/vendor/faiss/faiss/utils/simdlib_neon.h +832 -0
- data/vendor/faiss/faiss/utils/utils.cpp +304 -287
- data/vendor/faiss/faiss/utils/utils.h +54 -49
- metadata +29 -4
@@ -0,0 +1,679 @@
|
|
1
|
+
/**
|
2
|
+
* Copyright (c) Facebook, Inc. and its affiliates.
|
3
|
+
*
|
4
|
+
* This source code is licensed under the MIT license found in the
|
5
|
+
* LICENSE file in the root directory of this source tree.
|
6
|
+
*/
|
7
|
+
|
8
|
+
// -*- c++ -*-
|
9
|
+
|
10
|
+
#include <faiss/impl/NSG.h>
|
11
|
+
|
12
|
+
#include <algorithm>
|
13
|
+
#include <memory>
|
14
|
+
#include <mutex>
|
15
|
+
#include <stack>
|
16
|
+
|
17
|
+
#include <faiss/impl/AuxIndexStructures.h>
|
18
|
+
|
19
|
+
namespace faiss {
|
20
|
+
|
21
|
+
namespace nsg {
|
22
|
+
|
23
|
+
namespace {
|
24
|
+
|
25
|
+
// It needs to be smaller than 0
|
26
|
+
constexpr int EMPTY_ID = -1;
|
27
|
+
|
28
|
+
/* Wrap the distance computer into one that negates the
|
29
|
+
distances. This makes supporting INNER_PRODUCE search easier */
|
30
|
+
|
31
|
+
struct NegativeDistanceComputer : DistanceComputer {
|
32
|
+
using idx_t = Index::idx_t;
|
33
|
+
|
34
|
+
/// owned by this
|
35
|
+
DistanceComputer* basedis;
|
36
|
+
|
37
|
+
explicit NegativeDistanceComputer(DistanceComputer* basedis)
|
38
|
+
: basedis(basedis) {}
|
39
|
+
|
40
|
+
void set_query(const float* x) override {
|
41
|
+
basedis->set_query(x);
|
42
|
+
}
|
43
|
+
|
44
|
+
/// compute distance of vector i to current query
|
45
|
+
float operator()(idx_t i) override {
|
46
|
+
return -(*basedis)(i);
|
47
|
+
}
|
48
|
+
|
49
|
+
/// compute distance between two stored vectors
|
50
|
+
float symmetric_dis(idx_t i, idx_t j) override {
|
51
|
+
return -basedis->symmetric_dis(i, j);
|
52
|
+
}
|
53
|
+
|
54
|
+
~NegativeDistanceComputer() override {
|
55
|
+
delete basedis;
|
56
|
+
}
|
57
|
+
};
|
58
|
+
|
59
|
+
} // namespace
|
60
|
+
|
61
|
+
DistanceComputer* storage_distance_computer(const Index* storage) {
|
62
|
+
if (storage->metric_type == METRIC_INNER_PRODUCT) {
|
63
|
+
return new NegativeDistanceComputer(storage->get_distance_computer());
|
64
|
+
} else {
|
65
|
+
return storage->get_distance_computer();
|
66
|
+
}
|
67
|
+
}
|
68
|
+
|
69
|
+
} // namespace nsg
|
70
|
+
|
71
|
+
using namespace nsg;
|
72
|
+
|
73
|
+
using LockGuard = std::lock_guard<std::mutex>;
|
74
|
+
|
75
|
+
struct Neighbor {
|
76
|
+
int id;
|
77
|
+
float distance;
|
78
|
+
bool flag;
|
79
|
+
|
80
|
+
Neighbor() = default;
|
81
|
+
Neighbor(int id, float distance, bool f)
|
82
|
+
: id(id), distance(distance), flag(f) {}
|
83
|
+
|
84
|
+
inline bool operator<(const Neighbor& other) const {
|
85
|
+
return distance < other.distance;
|
86
|
+
}
|
87
|
+
};
|
88
|
+
|
89
|
+
struct Node {
|
90
|
+
int id;
|
91
|
+
float distance;
|
92
|
+
|
93
|
+
Node() = default;
|
94
|
+
Node(int id, float distance) : id(id), distance(distance) {}
|
95
|
+
|
96
|
+
inline bool operator<(const Node& other) const {
|
97
|
+
return distance < other.distance;
|
98
|
+
}
|
99
|
+
};
|
100
|
+
|
101
|
+
inline int insert_into_pool(Neighbor* addr, int K, Neighbor nn) {
|
102
|
+
// find the location to insert
|
103
|
+
int left = 0, right = K - 1;
|
104
|
+
if (addr[left].distance > nn.distance) {
|
105
|
+
memmove(&addr[left + 1], &addr[left], K * sizeof(Neighbor));
|
106
|
+
addr[left] = nn;
|
107
|
+
return left;
|
108
|
+
}
|
109
|
+
if (addr[right].distance < nn.distance) {
|
110
|
+
addr[K] = nn;
|
111
|
+
return K;
|
112
|
+
}
|
113
|
+
while (left < right - 1) {
|
114
|
+
int mid = (left + right) / 2;
|
115
|
+
if (addr[mid].distance > nn.distance) {
|
116
|
+
right = mid;
|
117
|
+
} else {
|
118
|
+
left = mid;
|
119
|
+
}
|
120
|
+
}
|
121
|
+
// check equal ID
|
122
|
+
|
123
|
+
while (left > 0) {
|
124
|
+
if (addr[left].distance < nn.distance) {
|
125
|
+
break;
|
126
|
+
}
|
127
|
+
if (addr[left].id == nn.id) {
|
128
|
+
return K + 1;
|
129
|
+
}
|
130
|
+
left--;
|
131
|
+
}
|
132
|
+
if (addr[left].id == nn.id || addr[right].id == nn.id) {
|
133
|
+
return K + 1;
|
134
|
+
}
|
135
|
+
memmove(&addr[right + 1], &addr[right], (K - right) * sizeof(Neighbor));
|
136
|
+
addr[right] = nn;
|
137
|
+
return right;
|
138
|
+
}
|
139
|
+
|
140
|
+
NSG::NSG(int R) : R(R), rng(0x0903) {
|
141
|
+
L = R + 32;
|
142
|
+
C = R + 100;
|
143
|
+
search_L = 16;
|
144
|
+
ntotal = 0;
|
145
|
+
is_built = false;
|
146
|
+
srand(0x1998);
|
147
|
+
}
|
148
|
+
|
149
|
+
void NSG::search(
|
150
|
+
DistanceComputer& dis,
|
151
|
+
int k,
|
152
|
+
idx_t* I,
|
153
|
+
float* D,
|
154
|
+
VisitedTable& vt) const {
|
155
|
+
FAISS_THROW_IF_NOT(is_built);
|
156
|
+
FAISS_THROW_IF_NOT(final_graph);
|
157
|
+
|
158
|
+
int pool_size = std::max(search_L, k);
|
159
|
+
std::vector<Neighbor> retset;
|
160
|
+
std::vector<Node> tmp;
|
161
|
+
search_on_graph<false>(
|
162
|
+
*final_graph, dis, vt, enterpoint, pool_size, retset, tmp);
|
163
|
+
|
164
|
+
for (size_t i = 0; i < k; i++) {
|
165
|
+
I[i] = retset[i].id;
|
166
|
+
D[i] = retset[i].distance;
|
167
|
+
}
|
168
|
+
}
|
169
|
+
|
170
|
+
void NSG::build(
|
171
|
+
Index* storage,
|
172
|
+
idx_t n,
|
173
|
+
const nsg::Graph<idx_t>& knn_graph,
|
174
|
+
bool verbose) {
|
175
|
+
FAISS_THROW_IF_NOT(!is_built && ntotal == 0);
|
176
|
+
|
177
|
+
if (verbose) {
|
178
|
+
printf("NSG::build R=%d, L=%d, C=%d\n", R, L, C);
|
179
|
+
}
|
180
|
+
|
181
|
+
ntotal = n;
|
182
|
+
init_graph(storage, knn_graph);
|
183
|
+
|
184
|
+
std::vector<int> degrees(n, 0);
|
185
|
+
{
|
186
|
+
nsg::Graph<Node> tmp_graph(n, R);
|
187
|
+
|
188
|
+
link(storage, knn_graph, tmp_graph, verbose);
|
189
|
+
|
190
|
+
final_graph = std::make_shared<nsg::Graph<int>>(n, R);
|
191
|
+
std::fill_n(final_graph->data, n * R, EMPTY_ID);
|
192
|
+
|
193
|
+
#pragma omp parallel for
|
194
|
+
for (int i = 0; i < n; i++) {
|
195
|
+
int cnt = 0;
|
196
|
+
for (int j = 0; j < R; j++) {
|
197
|
+
int id = tmp_graph.at(i, j).id;
|
198
|
+
if (id != EMPTY_ID) {
|
199
|
+
final_graph->at(i, cnt) = id;
|
200
|
+
cnt += 1;
|
201
|
+
}
|
202
|
+
degrees[i] = cnt;
|
203
|
+
}
|
204
|
+
}
|
205
|
+
}
|
206
|
+
|
207
|
+
int num_attached = tree_grow(storage, degrees);
|
208
|
+
check_graph();
|
209
|
+
is_built = true;
|
210
|
+
|
211
|
+
if (verbose) {
|
212
|
+
int max = 0, min = 1e6;
|
213
|
+
double avg = 0;
|
214
|
+
|
215
|
+
for (int i = 0; i < n; i++) {
|
216
|
+
int size = 0;
|
217
|
+
while (size < R && final_graph->at(i, size) != EMPTY_ID) {
|
218
|
+
size += 1;
|
219
|
+
}
|
220
|
+
max = std::max(size, max);
|
221
|
+
min = std::min(size, min);
|
222
|
+
avg += size;
|
223
|
+
}
|
224
|
+
|
225
|
+
avg = avg / n;
|
226
|
+
printf("Degree Statistics: Max = %d, Min = %d, Avg = %lf\n",
|
227
|
+
max,
|
228
|
+
min,
|
229
|
+
avg);
|
230
|
+
printf("Attached nodes: %d\n", num_attached);
|
231
|
+
}
|
232
|
+
}
|
233
|
+
|
234
|
+
void NSG::reset() {
|
235
|
+
final_graph.reset();
|
236
|
+
ntotal = 0;
|
237
|
+
is_built = false;
|
238
|
+
}
|
239
|
+
|
240
|
+
void NSG::init_graph(Index* storage, const nsg::Graph<idx_t>& knn_graph) {
|
241
|
+
int d = storage->d;
|
242
|
+
int n = storage->ntotal;
|
243
|
+
|
244
|
+
std::unique_ptr<float[]> center(new float[d]);
|
245
|
+
std::unique_ptr<float[]> tmp(new float[d]);
|
246
|
+
std::fill_n(center.get(), d, 0.0f);
|
247
|
+
|
248
|
+
for (int i = 0; i < n; i++) {
|
249
|
+
storage->reconstruct(i, tmp.get());
|
250
|
+
for (int j = 0; j < d; j++) {
|
251
|
+
center[j] += tmp[j];
|
252
|
+
}
|
253
|
+
}
|
254
|
+
|
255
|
+
for (int i = 0; i < d; i++) {
|
256
|
+
center[i] /= n;
|
257
|
+
}
|
258
|
+
|
259
|
+
std::vector<Neighbor> retset;
|
260
|
+
std::vector<Node> tmpset;
|
261
|
+
|
262
|
+
// random initialize navigating point
|
263
|
+
int ep = rng.rand_int(n);
|
264
|
+
std::unique_ptr<DistanceComputer> dis(storage_distance_computer(storage));
|
265
|
+
|
266
|
+
dis->set_query(center.get());
|
267
|
+
VisitedTable vt(ntotal);
|
268
|
+
|
269
|
+
// Do not collect the visited nodes
|
270
|
+
search_on_graph<false>(knn_graph, *dis, vt, ep, L, retset, tmpset);
|
271
|
+
|
272
|
+
// set enterpoint
|
273
|
+
enterpoint = retset[0].id;
|
274
|
+
}
|
275
|
+
|
276
|
+
template <bool collect_fullset, class index_t>
|
277
|
+
void NSG::search_on_graph(
|
278
|
+
const nsg::Graph<index_t>& graph,
|
279
|
+
DistanceComputer& dis,
|
280
|
+
VisitedTable& vt,
|
281
|
+
int ep,
|
282
|
+
int pool_size,
|
283
|
+
std::vector<Neighbor>& retset,
|
284
|
+
std::vector<Node>& fullset) const {
|
285
|
+
RandomGenerator gen(0x1234);
|
286
|
+
retset.resize(pool_size + 1);
|
287
|
+
std::vector<int> init_ids(pool_size);
|
288
|
+
|
289
|
+
int num_ids = 0;
|
290
|
+
for (int i = 0; i < init_ids.size() && i < graph.K; i++) {
|
291
|
+
int id = (int)graph.at(ep, i);
|
292
|
+
if (id < 0 || id >= ntotal) {
|
293
|
+
continue;
|
294
|
+
}
|
295
|
+
|
296
|
+
init_ids[i] = id;
|
297
|
+
vt.set(id);
|
298
|
+
num_ids += 1;
|
299
|
+
}
|
300
|
+
|
301
|
+
while (num_ids < pool_size) {
|
302
|
+
int id = gen.rand_int(ntotal);
|
303
|
+
if (vt.get(id)) {
|
304
|
+
continue;
|
305
|
+
}
|
306
|
+
|
307
|
+
init_ids[num_ids] = id;
|
308
|
+
num_ids++;
|
309
|
+
vt.set(id);
|
310
|
+
}
|
311
|
+
|
312
|
+
for (int i = 0; i < init_ids.size(); i++) {
|
313
|
+
int id = init_ids[i];
|
314
|
+
|
315
|
+
float dist = dis(id);
|
316
|
+
retset[i] = Neighbor(id, dist, true);
|
317
|
+
|
318
|
+
if (collect_fullset) {
|
319
|
+
fullset.emplace_back(retset[i].id, retset[i].distance);
|
320
|
+
}
|
321
|
+
}
|
322
|
+
|
323
|
+
std::sort(retset.begin(), retset.begin() + pool_size);
|
324
|
+
|
325
|
+
int k = 0;
|
326
|
+
while (k < pool_size) {
|
327
|
+
int updated_pos = pool_size;
|
328
|
+
|
329
|
+
if (retset[k].flag) {
|
330
|
+
retset[k].flag = false;
|
331
|
+
int n = retset[k].id;
|
332
|
+
|
333
|
+
for (int m = 0; m < graph.K; m++) {
|
334
|
+
int id = (int)graph.at(n, m);
|
335
|
+
if (id < 0 || id > ntotal || vt.get(id)) {
|
336
|
+
continue;
|
337
|
+
}
|
338
|
+
vt.set(id);
|
339
|
+
|
340
|
+
float dist = dis(id);
|
341
|
+
Neighbor nn(id, dist, true);
|
342
|
+
if (collect_fullset) {
|
343
|
+
fullset.emplace_back(id, dist);
|
344
|
+
}
|
345
|
+
|
346
|
+
if (dist >= retset[pool_size - 1].distance) {
|
347
|
+
continue;
|
348
|
+
}
|
349
|
+
|
350
|
+
int r = insert_into_pool(retset.data(), pool_size, nn);
|
351
|
+
|
352
|
+
updated_pos = std::min(updated_pos, r);
|
353
|
+
}
|
354
|
+
}
|
355
|
+
|
356
|
+
k = (updated_pos <= k) ? updated_pos : (k + 1);
|
357
|
+
}
|
358
|
+
}
|
359
|
+
|
360
|
+
void NSG::link(
|
361
|
+
Index* storage,
|
362
|
+
const nsg::Graph<idx_t>& knn_graph,
|
363
|
+
nsg::Graph<Node>& graph,
|
364
|
+
bool /* verbose */) {
|
365
|
+
#pragma omp parallel
|
366
|
+
{
|
367
|
+
std::unique_ptr<float[]> vec(new float[storage->d]);
|
368
|
+
|
369
|
+
std::vector<Node> pool;
|
370
|
+
std::vector<Neighbor> tmp;
|
371
|
+
|
372
|
+
VisitedTable vt(ntotal);
|
373
|
+
std::unique_ptr<DistanceComputer> dis(
|
374
|
+
storage_distance_computer(storage));
|
375
|
+
|
376
|
+
#pragma omp for schedule(dynamic, 100)
|
377
|
+
for (int i = 0; i < ntotal; i++) {
|
378
|
+
storage->reconstruct(i, vec.get());
|
379
|
+
dis->set_query(vec.get());
|
380
|
+
|
381
|
+
// Collect the visited nodes into pool
|
382
|
+
search_on_graph<true>(
|
383
|
+
knn_graph, *dis, vt, enterpoint, L, tmp, pool);
|
384
|
+
|
385
|
+
sync_prune(i, pool, *dis, vt, knn_graph, graph);
|
386
|
+
|
387
|
+
pool.clear();
|
388
|
+
tmp.clear();
|
389
|
+
vt.advance();
|
390
|
+
}
|
391
|
+
} // omp parallel
|
392
|
+
|
393
|
+
std::vector<std::mutex> locks(ntotal);
|
394
|
+
#pragma omp parallel
|
395
|
+
{
|
396
|
+
std::unique_ptr<DistanceComputer> dis(
|
397
|
+
storage_distance_computer(storage));
|
398
|
+
|
399
|
+
#pragma omp for schedule(dynamic, 100)
|
400
|
+
for (int i = 0; i < ntotal; ++i) {
|
401
|
+
add_reverse_links(i, locks, *dis, graph);
|
402
|
+
}
|
403
|
+
} // omp parallel
|
404
|
+
}
|
405
|
+
|
406
|
+
void NSG::sync_prune(
|
407
|
+
int q,
|
408
|
+
std::vector<Node>& pool,
|
409
|
+
DistanceComputer& dis,
|
410
|
+
VisitedTable& vt,
|
411
|
+
const nsg::Graph<idx_t>& knn_graph,
|
412
|
+
nsg::Graph<Node>& graph) {
|
413
|
+
for (int i = 0; i < knn_graph.K; i++) {
|
414
|
+
int id = knn_graph.at(q, i);
|
415
|
+
if (id < 0 || id >= ntotal || vt.get(id)) {
|
416
|
+
continue;
|
417
|
+
}
|
418
|
+
|
419
|
+
float dist = dis.symmetric_dis(q, id);
|
420
|
+
pool.emplace_back(id, dist);
|
421
|
+
}
|
422
|
+
|
423
|
+
std::sort(pool.begin(), pool.end());
|
424
|
+
|
425
|
+
std::vector<Node> result;
|
426
|
+
|
427
|
+
int start = 0;
|
428
|
+
if (pool[start].id == q) {
|
429
|
+
start++;
|
430
|
+
}
|
431
|
+
result.push_back(pool[start]);
|
432
|
+
|
433
|
+
while (result.size() < R && (++start) < pool.size() && start < C) {
|
434
|
+
auto& p = pool[start];
|
435
|
+
bool occlude = false;
|
436
|
+
for (int t = 0; t < result.size(); t++) {
|
437
|
+
if (p.id == result[t].id) {
|
438
|
+
occlude = true;
|
439
|
+
break;
|
440
|
+
}
|
441
|
+
float djk = dis.symmetric_dis(result[t].id, p.id);
|
442
|
+
if (djk < p.distance /* dik */) {
|
443
|
+
occlude = true;
|
444
|
+
break;
|
445
|
+
}
|
446
|
+
}
|
447
|
+
if (!occlude) {
|
448
|
+
result.push_back(p);
|
449
|
+
}
|
450
|
+
}
|
451
|
+
|
452
|
+
for (size_t i = 0; i < R; i++) {
|
453
|
+
if (i < result.size()) {
|
454
|
+
graph.at(q, i).id = result[i].id;
|
455
|
+
graph.at(q, i).distance = result[i].distance;
|
456
|
+
} else {
|
457
|
+
graph.at(q, i).id = EMPTY_ID;
|
458
|
+
}
|
459
|
+
}
|
460
|
+
}
|
461
|
+
|
462
|
+
void NSG::add_reverse_links(
|
463
|
+
int q,
|
464
|
+
std::vector<std::mutex>& locks,
|
465
|
+
DistanceComputer& dis,
|
466
|
+
nsg::Graph<Node>& graph) {
|
467
|
+
for (size_t i = 0; i < R; i++) {
|
468
|
+
if (graph.at(q, i).id == EMPTY_ID) {
|
469
|
+
break;
|
470
|
+
}
|
471
|
+
|
472
|
+
Node sn(q, graph.at(q, i).distance);
|
473
|
+
int des = graph.at(q, i).id;
|
474
|
+
|
475
|
+
std::vector<Node> tmp_pool;
|
476
|
+
int dup = 0;
|
477
|
+
{
|
478
|
+
LockGuard guard(locks[des]);
|
479
|
+
for (int j = 0; j < R; j++) {
|
480
|
+
if (graph.at(des, j).id == EMPTY_ID) {
|
481
|
+
break;
|
482
|
+
}
|
483
|
+
if (q == graph.at(des, j).id) {
|
484
|
+
dup = 1;
|
485
|
+
break;
|
486
|
+
}
|
487
|
+
tmp_pool.push_back(graph.at(des, j));
|
488
|
+
}
|
489
|
+
}
|
490
|
+
|
491
|
+
if (dup) {
|
492
|
+
continue;
|
493
|
+
}
|
494
|
+
|
495
|
+
tmp_pool.push_back(sn);
|
496
|
+
if (tmp_pool.size() > R) {
|
497
|
+
std::vector<Node> result;
|
498
|
+
int start = 0;
|
499
|
+
std::sort(tmp_pool.begin(), tmp_pool.end());
|
500
|
+
result.push_back(tmp_pool[start]);
|
501
|
+
|
502
|
+
while (result.size() < R && (++start) < tmp_pool.size()) {
|
503
|
+
auto& p = tmp_pool[start];
|
504
|
+
bool occlude = false;
|
505
|
+
|
506
|
+
for (int t = 0; t < result.size(); t++) {
|
507
|
+
if (p.id == result[t].id) {
|
508
|
+
occlude = true;
|
509
|
+
break;
|
510
|
+
}
|
511
|
+
float djk = dis.symmetric_dis(result[t].id, p.id);
|
512
|
+
if (djk < p.distance /* dik */) {
|
513
|
+
occlude = true;
|
514
|
+
break;
|
515
|
+
}
|
516
|
+
}
|
517
|
+
|
518
|
+
if (!occlude) {
|
519
|
+
result.push_back(p);
|
520
|
+
}
|
521
|
+
}
|
522
|
+
|
523
|
+
{
|
524
|
+
LockGuard guard(locks[des]);
|
525
|
+
for (int t = 0; t < result.size(); t++) {
|
526
|
+
graph.at(des, t) = result[t];
|
527
|
+
}
|
528
|
+
}
|
529
|
+
|
530
|
+
} else {
|
531
|
+
LockGuard guard(locks[des]);
|
532
|
+
for (int t = 0; t < R; t++) {
|
533
|
+
if (graph.at(des, t).id == EMPTY_ID) {
|
534
|
+
graph.at(des, t) = sn;
|
535
|
+
break;
|
536
|
+
}
|
537
|
+
}
|
538
|
+
}
|
539
|
+
}
|
540
|
+
}
|
541
|
+
|
542
|
+
int NSG::tree_grow(Index* storage, std::vector<int>& degrees) {
|
543
|
+
int root = enterpoint;
|
544
|
+
VisitedTable vt(ntotal);
|
545
|
+
VisitedTable vt2(ntotal);
|
546
|
+
|
547
|
+
int num_attached = 0;
|
548
|
+
int cnt = 0;
|
549
|
+
while (true) {
|
550
|
+
cnt = dfs(vt, root, cnt);
|
551
|
+
if (cnt >= ntotal) {
|
552
|
+
break;
|
553
|
+
}
|
554
|
+
|
555
|
+
root = attach_unlinked(storage, vt, vt2, degrees);
|
556
|
+
vt2.advance();
|
557
|
+
num_attached += 1;
|
558
|
+
}
|
559
|
+
|
560
|
+
return num_attached;
|
561
|
+
}
|
562
|
+
|
563
|
+
int NSG::dfs(VisitedTable& vt, int root, int cnt) const {
|
564
|
+
int node = root;
|
565
|
+
std::stack<int> stack;
|
566
|
+
stack.push(root);
|
567
|
+
|
568
|
+
if (!vt.get(root)) {
|
569
|
+
cnt++;
|
570
|
+
}
|
571
|
+
vt.set(root);
|
572
|
+
|
573
|
+
while (!stack.empty()) {
|
574
|
+
int next = EMPTY_ID;
|
575
|
+
for (int i = 0; i < R; i++) {
|
576
|
+
int id = final_graph->at(node, i);
|
577
|
+
if (id != EMPTY_ID && !vt.get(id)) {
|
578
|
+
next = id;
|
579
|
+
break;
|
580
|
+
}
|
581
|
+
}
|
582
|
+
|
583
|
+
if (next == EMPTY_ID) {
|
584
|
+
stack.pop();
|
585
|
+
if (stack.empty()) {
|
586
|
+
break;
|
587
|
+
}
|
588
|
+
node = stack.top();
|
589
|
+
continue;
|
590
|
+
}
|
591
|
+
node = next;
|
592
|
+
vt.set(node);
|
593
|
+
stack.push(node);
|
594
|
+
cnt++;
|
595
|
+
}
|
596
|
+
|
597
|
+
return cnt;
|
598
|
+
}
|
599
|
+
|
600
|
+
int NSG::attach_unlinked(
|
601
|
+
Index* storage,
|
602
|
+
VisitedTable& vt,
|
603
|
+
VisitedTable& vt2,
|
604
|
+
std::vector<int>& degrees) {
|
605
|
+
/* NOTE: This implementation is slightly different from the original paper.
|
606
|
+
*
|
607
|
+
* Instead of connecting the unlinked node to the nearest point in the
|
608
|
+
* spanning tree which will increase the maximum degree of the graph and
|
609
|
+
* also make the graph hard to maintain, this implementation links the
|
610
|
+
* unlinked node to the nearest node of which the degree is smaller than R.
|
611
|
+
* It will keep the degree of all nodes to be no more than `R`.
|
612
|
+
*/
|
613
|
+
|
614
|
+
// find one unlinked node
|
615
|
+
int id = EMPTY_ID;
|
616
|
+
for (int i = 0; i < ntotal; i++) {
|
617
|
+
if (!vt.get(i)) {
|
618
|
+
id = i;
|
619
|
+
break;
|
620
|
+
}
|
621
|
+
}
|
622
|
+
|
623
|
+
if (id == EMPTY_ID) {
|
624
|
+
return EMPTY_ID; // No Unlinked Node
|
625
|
+
}
|
626
|
+
|
627
|
+
std::vector<Neighbor> tmp;
|
628
|
+
std::vector<Node> pool;
|
629
|
+
|
630
|
+
std::unique_ptr<DistanceComputer> dis(storage_distance_computer(storage));
|
631
|
+
std::unique_ptr<float[]> vec(new float[storage->d]);
|
632
|
+
|
633
|
+
storage->reconstruct(id, vec.get());
|
634
|
+
dis->set_query(vec.get());
|
635
|
+
|
636
|
+
// Collect the visited nodes into pool
|
637
|
+
search_on_graph<true>(
|
638
|
+
*final_graph, *dis, vt2, enterpoint, search_L, tmp, pool);
|
639
|
+
|
640
|
+
std::sort(pool.begin(), pool.end());
|
641
|
+
|
642
|
+
int node;
|
643
|
+
bool found = false;
|
644
|
+
for (int i = 0; i < pool.size(); i++) {
|
645
|
+
node = pool[i].id;
|
646
|
+
if (degrees[node] < R && node != id) {
|
647
|
+
found = true;
|
648
|
+
break;
|
649
|
+
}
|
650
|
+
}
|
651
|
+
|
652
|
+
// randomly choice annother node
|
653
|
+
if (!found) {
|
654
|
+
do {
|
655
|
+
node = rng.rand_int(ntotal);
|
656
|
+
if (vt.get(node) && degrees[node] < R && node != id) {
|
657
|
+
found = true;
|
658
|
+
}
|
659
|
+
} while (!found);
|
660
|
+
}
|
661
|
+
|
662
|
+
int pos = degrees[node];
|
663
|
+
final_graph->at(node, pos) = id; // replace
|
664
|
+
degrees[node] += 1;
|
665
|
+
|
666
|
+
return node;
|
667
|
+
}
|
668
|
+
|
669
|
+
void NSG::check_graph() const {
|
670
|
+
#pragma omp parallel for
|
671
|
+
for (int i = 0; i < ntotal; i++) {
|
672
|
+
for (int j = 0; j < R; j++) {
|
673
|
+
int id = final_graph->at(i, j);
|
674
|
+
FAISS_THROW_IF_NOT(id < ntotal && (id >= 0 || id == EMPTY_ID));
|
675
|
+
}
|
676
|
+
}
|
677
|
+
}
|
678
|
+
|
679
|
+
} // namespace faiss
|