mirror of
https://github.com/facebookresearch/faiss.git
synced 2025-06-03 21:54:02 +08:00
a8118acbc5
Changelog: - changed license: BSD+Patents -> MIT - propagates exceptions raised in sub-indexes of IndexShards and IndexReplicas - support for searching several inverted lists in parallel (parallel_mode != 0) - better support for PQ codes where nbit != 8 or 16 - IVFSpectralHash implementation: spectral hash codes inside an IVF - 6-bit per component scalar quantizer (4 and 8 bit were already supported) - combinations of inverted lists: HStackInvertedLists and VStackInvertedLists - configurable number of threads for OnDiskInvertedLists prefetching (including 0=no prefetch) - more test and demo code compatible with Python 3 (print with parentheses) - refactored benchmark code: data loading is now in a single file
67 lines
1.6 KiB
C++
67 lines
1.6 KiB
C++
/**
|
|
* Copyright (c) Facebook, Inc. and its affiliates.
|
|
*
|
|
* This source code is licensed under the MIT license found in the
|
|
* LICENSE file in the root directory of this source tree.
|
|
*/
|
|
|
|
#include <cstdio>
|
|
#include <cstdlib>
|
|
|
|
#include <gtest/gtest.h>
|
|
|
|
#include <faiss/IndexIVFPQ.h>
|
|
#include <faiss/IndexFlat.h>
|
|
#include <faiss/utils.h>
|
|
|
|
|
|
namespace {
|
|
|
|
// dimension of the vectors to index
|
|
int d = 64;
|
|
|
|
// size of the database we plan to index
|
|
size_t nb = 8000;
|
|
|
|
|
|
double eval_codec_error (long ncentroids, long m, const std::vector<float> &v)
|
|
{
|
|
faiss::IndexFlatL2 coarse_quantizer (d);
|
|
faiss::IndexIVFPQ index (&coarse_quantizer, d,
|
|
ncentroids, m, 8);
|
|
index.pq.cp.niter = 10; // speed up train
|
|
index.train (nb, v.data());
|
|
|
|
// encode and decode to compute reconstruction error
|
|
|
|
std::vector<long> keys (nb);
|
|
std::vector<uint8_t> codes (nb * m);
|
|
index.encode_multiple (nb, keys.data(), v.data(), codes.data(), true);
|
|
|
|
std::vector<float> v2 (nb * d);
|
|
index.decode_multiple (nb, keys.data(), codes.data(), v2.data());
|
|
|
|
return faiss::fvec_L2sqr (v.data(), v2.data(), nb * d);
|
|
}
|
|
|
|
} // namespace
|
|
|
|
|
|
TEST(IVFPQ, codec) {
|
|
|
|
std::vector <float> database (nb * d);
|
|
for (size_t i = 0; i < nb * d; i++) {
|
|
database[i] = drand48();
|
|
}
|
|
|
|
double err0 = eval_codec_error(16, 8, database);
|
|
|
|
// should be more accurate as there are more coarse centroids
|
|
double err1 = eval_codec_error(128, 8, database);
|
|
EXPECT_GT(err0, err1);
|
|
|
|
// should be more accurate as there are more PQ codes
|
|
double err2 = eval_codec_error(16, 16, database);
|
|
EXPECT_GT(err0, err2);
|
|
}
|