Skip to content

v0.17.0

Choose a tag to compare

@wxyucs wxyucs released this 13 Oct 03:58
787d503

What's Changed

Bug Fixes

  • bugfix on fast bitset's AND operator by @LHT129 in #1036
  • upgrade robin_map version to v0.14.0 by @inabao in #1040
  • fix qps reduce in sindi by @inabao in #1048
  • fix incorrect CXX ABI in binary distribution by @wxyucs in #1044
  • fix compile error in tests by @inabao in #1058
  • fix(sindi): fix the issue where Sindi returns irrelevant results by @inabao in #1085
  • fix(hgraph): fix HGraph deserialization failure by @inabao in #1083
  • fix(sindi): fix compilation issues of sindi in versions prior to C++17 by @inabao in #1080
  • fix(sindi): fix redundant results in the return data by @inabao in #1108
  • fix: add missing include in commen.h by @wxyucs in #1127
  • fix(deserialize): fix core when deserializing empty index by @inabao in #1150
  • fix(index): add check for empty input in index by @inabao in #1159
  • fix(int8): fix the missing header files issue by @inabao in #1167
  • fix(neon): asan problem on load data by @LHT129 in #1174
  • fix(hnsw): performance degradation after remove by @ShawnShawnYou in #1176
  • fix(serialize): support empty serialize for export model by @LHT129 in #1169
  • fix(hgraph): fix recall issues in HGraph merge by @inabao in #1189
  • fix(cmake): fix wrong message on RUNTIME_AVX512VPOPCNTDQ by @LHT129 in #1198
  • fix: LabelTable deserialzie error by @LHT129 in #1208
  • fix(quantizer): modify the instantiation location of the adapter by @inabao in #1211
  • fix(dataset): fix the string deep copy not being properly implemented by @inabao in #1214
  • fix(attribute): fix attribute init with use_bucket by @LHT129 in #1216
  • fix(hnsw): performance degradation by @ShawnShawnYou in #1222
  • fix(attribute): fix loss json init for bruteforce by @LHT129 in #1220

New Features

  • add overtime check on search by @LHT129 in #1027
  • Introduce transform quantizer by @ShawnShawnYou in #1031
  • Support Export All ID from index by @LHT129 in #1069
  • support tombstone recovery in hnsw by @ShawnShawnYou in #1056
  • Support data export on IVF Index by @LHT129 in #1052
  • feat(ivf): support calculate distance by id for IVF by @inabao in #1096
  • feat(dataset): support Append and DeepCopy methods in Dataset by @inabao in #1074
  • feat(index): support GetDataByIdWithFlag by @LHT129 in #1105
  • feat(search): extend search request to cover all search cases by @LHT129 in #1123
  • feat(extra info): Support ExtraInfo for all index by @LHT129 in #1135
  • feat(HGraph): Support Hgraph's graph storage with different io config by @LHT129 in #1136
  • feat(HGraph): add multi-threaded searcher to parallelize single query execution by @mly5269 in #1071
  • feat(hgraph): support set topk factor while searching by @inabao in #1180
  • feat(sindi): support passing an allocator into KnnSearch by @inabao in #1195

Improvements

  • refactor heap merge on ivf parallel search by @LHT129 in #1012
  • strip release of pyvsag by @LHT129 in #1021
  • add more tag index check on compatibility by @LHT129 in #1024
  • add benchmark parameters of popular datasets by @wxyucs in #1005
  • add tool to analyze index for hgraph by @inabao in #1025
  • increase beam width during reordering by @inabao in #1037
  • add file remove check on disk io type by @LHT129 in #1023
  • reduce memory on fast_bitset by @LHT129 in #1042
  • replace Vector with raw pointer on bitset by @LHT129 in #1067
  • support store raw vector on HGraph by @LHT129 in #1063
  • support analyze HGraph by searching by @inabao in #1045
  • refactor(quantizer): optimize quantizer's compute by @LHT129 in #1089
  • feat(reader): add the MultRead method to the reader interface by @inabao in #1082
  • refactor: declare shared_ptr/unique_ptr with marco by @LHT129 in #1095
  • support in-place update in HGraph by @ShawnShawnYou in #1049
  • chore: add more tag for compatibility by @LHT129 in #1101
  • refactor(index): reorder the huge index's function by @LHT129 in #1103
  • feat(sindi): support several interfaces by @ShawnShawnYou in #1111
  • feat(ivf): support for self-checking and query-based checking of IVF inverted indexes by @inabao in #1116
  • feat(SINDI): support batch distance calculate by @ShawnShawnYou in #1131
  • fix(SINDI): filter performance by @ShawnShawnYou in #1133
  • refactor(sindi): remove the dimension check for the sparse vector index by @inabao in #1130
  • feat(sindi): support exclude the footer during deserialize by @inabao in #1122
  • refactor(headers): improve header include rules by @LHT129 in #1118
  • feat(SINDI): support set immutable by @ShawnShawnYou in #1132
  • feat(simd): add int8 ip distance impl by @Coien-rr in #1100
  • feat(SINDI): support get memory usage by @ShawnShawnYou in #1141
  • typo(test): replace "[]" with "()" for catch2 test by @LHT129 in #1143
  • feat(sindi): add parameter check for sindi by @inabao in #1144
  • typo(index): typo error on index's serialize log by @LHT129 in #1139
  • refactor(test): reduce full test size by @LHT129 in #1138
  • refactor: use unique_ptr instead shared_ptr by @LHT129 in #1142
  • coverage: improve coverage rate by @LHT129 in #1117
  • refactor(kmeans): split kmeans compile by @LHT129 in #1137
  • refactor: move files by @LHT129 in #1158
  • feat(int8): add int8 type support for full-precision quantizer by @Coien-rr in #1152
  • refactor(hgraph): simplify hgraph's sparse parameter parsing logic by @inabao in #1164
  • feat(HGraph): support tombstone recovery by @ShawnShawnYou in #1160
  • refactor(footer): reduce io count when parsing footer by @inabao in #1171
  • refactor(option): lower the minimum block size limit by @inabao in #1165
  • feat(Hgraph): support tunning alpha in MRNG by @HeHuMing in #1104
  • refactor: json optimize compile by @LHT129 in #1168
  • refactor(test): clear ci log and cout test name by @inabao in #1182
  • tests: speed up functests by reduce dim upperbound by @LHT129 in #1187
  • refactor: rename data_cell to datacell by @wxyucs in #1178
  • chore(cmake): clear CMake warnings, add license to the file header by @wxyucs in #1193
  • refactor(attr): use parameter to create attribute datacell by @LHT129 in #1200
  • refactor: split unittest compile to speedup link by @LHT129 in #1202
  • feat(quantization): Introduce QuantizerAdapter to support non-FP32 data types in quantizers by @Coien-rr in #1185
  • tests: add serialize check on export model test by @LHT129 in #1203
  • feat: support use old serial format by @wxyucs in #1201
  • tests: relax recall threshold to fit random dataset by @LHT129 in #1204

Other Changes

  • update ivf doc by @LHT129 in #1011
  • fix label_remap size serialize by @shadowao in #1022
  • [skip ci] update document of hgraph by @inabao in #1032
  • update perf.yml for COHERE by @inabao in #1038
  • support sve to accelerate distance compute by @jac0626 in #989
  • support mrng in hgraph by @HeHuMing in #1079
  • [skip ci] docs(hgraph): refactor HGraph doc by @LHT129 in #1102
  • test(ivf): refactor ivf fp16 test on cosine metric by @LHT129 in #1106
  • docs(ivf): add parameter table for ivf doc by @LHT129 in #1120
  • [skip ci] docs(brute_force): add brute_force doc by @LHT129 in #1121
  • complete missing neon fuctions by @jac0626 in #1018
  • feat(sparse): improve performance of batch distances calculation by @ShawnShawnYou in #1192
  • [skip ci] doc: modify the formatting of publications and references by @inabao in #1207
  • refactor(reader_io): support reader_io omits I/O operations during deserialization by @inabao in #1212

New Contributors

Full Changelog: v0.16.0...v0.17.0