github alibaba/zvec v0.8.0-alpha

pre-release3 hours ago

What's Changed

  • feat(tools): support turbo quantizer in core tool set by @richyreachy in #684
  • perf: batch convert docs to achieve acceleration by @zzlin237 in #647
  • feat: add Windows ARM64 (MSVC) build support by @Aanerud in #352
  • chore: update README to v0.7.0 by @chinaux in #697
  • feat(refine): support configurable flat storage types by @luoxiaojian in #677
  • perf: improve CRC32C and FP16 runtime dispatch by @richyreachy in #700
  • fix(build): fix /Z7 /Zi conflict in Debug on Windows by @JalinWang in #692
  • refactor(turbo): add AVX2/AVX512 SIMD kernels for distance by @richyreachy in #687
  • feat(turbo): add PQ int4 quantizers and PQFast FastScan support by @zzlin237 in #664
  • feat(diskann): Add Windows DiskANN support with asynchronous I/O by @iaojnh in #625
  • feat(turbo): add AVX-512 FP16 distance kernels by @richyreachy in #710
  • refactor(turbo): migrate neon distance kernels by @JalinWang in #689
  • feat:support macOS x64 builds by @Cuiyus in #713
  • feat(quantization): add uniform uint4 reformer/quantizer by @JalinWang in #698
  • feat(turbo): add SSE distance kernels by @richyreachy in #718
  • perf(vamana): use split contiguous packed layout by @luoxiaojian in #701
  • fix: suppress build warning by @richyreachy in #720
  • fix: synchronize flat cache and collection block switching by @richyreachy in #726
  • fix(db): synchronize reads of writing segment by @iaojnh in #722
  • refactor(diskann): adapt diskann to turbo quantizer by @richyreachy in #678
  • fix(core): IVF n_iters propagation to clusterer by @JalinWang in #730
  • perf(vamana): port search and prefetch optimizations by @luoxiaojian in #727
  • ci(deps): bump actions/setup-python from 6 to 7 by @dependabot[bot] in #712
  • feat(diskann): add Android and iOS support with portable pread I/O by @iaojnh in #656
  • fix(sqlengine): balance deep OR expression trees by @Cuiyus in #694
  • fix: preserve vector row mapping when rebuilding indexes with deleted rows by @zhourrr in #731
  • perf(vamana): optimize search state and fix medoid selection by @luoxiaojian in #738
  • feat(buffer): harden shared buffer pool core by @iaojnh in #708
  • fix(db): SIGSEGV when fetch/query race a segment switch; unblock reader concurrency by @YongqiYin in #715
  • perf(refine): optimize shared kernels and candidate handoff by @luoxiaojian in #740
  • fix(arm): enable neon fp16 kernels via march auto-detect and runtime dispatch by @richyreachy in #723
  • chore: enable modernize-use-equals-default and modernize-use-equals-delete by @egolearner in #734
  • fix: prevent overlapping prefix and suffix matches in LIKE by @egolearner in #745
  • ci(deps): bump softprops/action-gh-release from 2 to 3 by @dependabot[bot] in #725
  • perf(uint4): optimize AVX-512 distance and fix Vamana medoid by @luoxiaojian in #749
  • perf(search): Optimize graph refine candidate export by @luoxiaojian in #748
  • ci(deps): bump actions/setup-java from 5.7.0 to 6.0.1 by @dependabot[bot] in #750
  • chore: enable modernize-redundant-void-arg check and fix existing violations by @egolearner in #751
  • fix(ci): skip unavailable legacy Android SDK tools package by @zhourrr in #753
  • fix: correct inverted-index LIKE suffix and empty-array queries by @zhourrr in #752
  • perf(ivf): reduce peak memory usage during index build and merge by @iaojnh in #733
  • feat(quantization): integrate uniform quantizers with DB and HNSW by @luoxiaojian in #754
  • refactor(flat): use turbo for distance calculation by @richyreachy in #716
  • fix(quantization): decode turbo sources during merge training by @iaojnh in #763
  • refactor(sqlengine): add extensible filter set rewriting by @egolearner in #690
  • feat(buffer): add buffer-backed storage and HNSW integration by @iaojnh in #742
  • feat(turbo): add OPQ preprocessor by @zzlin237 in #717
  • fix(reducer): reuse optimize thread pool for builders by @JalinWang in #757
  • feat: relax name validation and harden recovery by @zhourrr in #756
  • refactor(turbo): share implementation across pq quantizers by @richyreachy in #765
  • chore: enable readability-identifier-naming check and fix violations by @egolearner in #764
  • feat(buffer): integrate IVF and DiskANN with shared page cache by @iaojnh in #766
  • fix(fts): persist sealed postings before dropping side column families by @egolearner in #759
  • perf(hnsw-rabitq): optimize HNSW RaBitQ index building by @JalinWang in #755
  • perf(diskann): reduce peak memory usage during index build and merge by @iaojnh in #762
  • refactor(sqlengine): separate filter validation and execution binding by @egolearner in #769
  • refactor(hnsw): use turbo for distance calculation by @richyreachy in #735
  • chore: add clang-tidy pre-commit hook by @Mordris in #571
  • fix(quantizer): use rounded INT4 values for scoring metadata by @richyreachy in #777
  • fix(core): honor optimize threads in nested training(rabitQ/DiskANN) by @JalinWang in #775
  • feat(rabitq): upgrade RaBitQ-Library to v0.3.8 and enable Windows x86_64 by @egolearner in #780
  • fix(test): avoid MSVC ARM64 ICE in uniform_uint8_metric_test by @JasonXuDeveloper in #795
  • refactor(ivf): use turbo for distance calculation by @richyreachy in #773
  • fix(core): read keys_map iterator before releasing shared lock by @smartworldarafath in #779
  • feat(search): add fast_query for dense read-only collections by @luoxiaojian in #761
  • fix(build): disable unused Thrift OpenSSL support by @egolearner in #819

New Contributors

Full Changelog: v0.7.0...v0.8.0-alpha

What's Changed

  • feat(tools): support turbo quantizer in core tool set by @richyreachy in #684
  • perf: batch convert docs to achieve acceleration by @zzlin237 in #647
  • feat: add Windows ARM64 (MSVC) build support by @Aanerud in #352
  • chore: update README to v0.7.0 by @chinaux in #697
  • feat(refine): support configurable flat storage types by @luoxiaojian in #677
  • perf: improve CRC32C and FP16 runtime dispatch by @richyreachy in #700
  • fix(build): fix /Z7 /Zi conflict in Debug on Windows by @JalinWang in #692
  • refactor(turbo): add AVX2/AVX512 SIMD kernels for distance by @richyreachy in #687
  • feat(turbo): add PQ int4 quantizers and PQFast FastScan support by @zzlin237 in #664
  • feat(diskann): Add Windows DiskANN support with asynchronous I/O by @iaojnh in #625
  • feat(turbo): add AVX-512 FP16 distance kernels by @richyreachy in #710
  • refactor(turbo): migrate neon distance kernels by @JalinWang in #689
  • feat:support macOS x64 builds by @Cuiyus in #713
  • feat(quantization): add uniform uint4 reformer/quantizer by @JalinWang in #698
  • feat(turbo): add SSE distance kernels by @richyreachy in #718
  • perf(vamana): use split contiguous packed layout by @luoxiaojian in #701
  • fix: suppress build warning by @richyreachy in #720
  • fix: synchronize flat cache and collection block switching by @richyreachy in #726
  • fix(db): synchronize reads of writing segment by @iaojnh in #722
  • refactor(diskann): adapt diskann to turbo quantizer by @richyreachy in #678
  • fix(core): IVF n_iters propagation to clusterer by @JalinWang in #730
  • perf(vamana): port search and prefetch optimizations by @luoxiaojian in #727
  • ci(deps): bump actions/setup-python from 6 to 7 by @dependabot[bot] in #712
  • feat(diskann): add Android and iOS support with portable pread I/O by @iaojnh in #656
  • fix(sqlengine): balance deep OR expression trees by @Cuiyus in #694
  • fix: preserve vector row mapping when rebuilding indexes with deleted rows by @zhourrr in #731
  • perf(vamana): optimize search state and fix medoid selection by @luoxiaojian in #738
  • feat(buffer): harden shared buffer pool core by @iaojnh in #708
  • fix(db): SIGSEGV when fetch/query race a segment switch; unblock reader concurrency by @YongqiYin in #715
  • perf(refine): optimize shared kernels and candidate handoff by @luoxiaojian in #740
  • fix(arm): enable neon fp16 kernels via march auto-detect and runtime dispatch by @richyreachy in #723
  • chore: enable modernize-use-equals-default and modernize-use-equals-delete by @egolearner in #734
  • fix: prevent overlapping prefix and suffix matches in LIKE by @egolearner in #745
  • ci(deps): bump softprops/action-gh-release from 2 to 3 by @dependabot[bot] in #725
  • perf(uint4): optimize AVX-512 distance and fix Vamana medoid by @luoxiaojian in #749
  • perf(search): Optimize graph refine candidate export by @luoxiaojian in #748
  • ci(deps): bump actions/setup-java from 5.7.0 to 6.0.1 by @dependabot[bot] in #750
  • chore: enable modernize-redundant-void-arg check and fix existing violations by @egolearner in #751
  • fix(ci): skip unavailable legacy Android SDK tools package by @zhourrr in #753
  • fix: correct inverted-index LIKE suffix and empty-array queries by @zhourrr in #752
  • perf(ivf): reduce peak memory usage during index build and merge by @iaojnh in #733
  • feat(quantization): integrate uniform quantizers with DB and HNSW by @luoxiaojian in #754
  • refactor(flat): use turbo for distance calculation by @richyreachy in #716
  • fix(quantization): decode turbo sources during merge training by @iaojnh in #763
  • refactor(sqlengine): add extensible filter set rewriting by @egolearner in #690
  • feat(buffer): add buffer-backed storage and HNSW integration by @iaojnh in #742
  • feat(turbo): add OPQ preprocessor by @zzlin237 in #717
  • fix(reducer): reuse optimize thread pool for builders by @JalinWang in #757
  • feat: relax name validation and harden recovery by @zhourrr in #756
  • refactor(turbo): share implementation across pq quantizers by @richyreachy in #765
  • chore: enable readability-identifier-naming check and fix violations by @egolearner in #764
  • feat(buffer): integrate IVF and DiskANN with shared page cache by @iaojnh in #766
  • fix(fts): persist sealed postings before dropping side column families by @egolearner in #759
  • perf(hnsw-rabitq): optimize HNSW RaBitQ index building by @JalinWang in #755
  • perf(diskann): reduce peak memory usage during index build and merge by @iaojnh in #762
  • refactor(sqlengine): separate filter validation and execution binding by @egolearner in #769
  • refactor(hnsw): use turbo for distance calculation by @richyreachy in #735
  • chore: add clang-tidy pre-commit hook by @Mordris in #571
  • fix(quantizer): use rounded INT4 values for scoring metadata by @richyreachy in #777
  • fix(core): honor optimize threads in nested training(rabitQ/DiskANN) by @JalinWang in #775
  • feat(rabitq): upgrade RaBitQ-Library to v0.3.8 and enable Windows x86_64 by @egolearner in #780
  • fix(test): avoid MSVC ARM64 ICE in uniform_uint8_metric_test by @JasonXuDeveloper in #795
  • refactor(ivf): use turbo for distance calculation by @richyreachy in #773
  • fix(core): read keys_map iterator before releasing shared lock by @smartworldarafath in #779
  • feat(search): add fast_query for dense read-only collections by @luoxiaojian in #761
  • fix(build): disable unused Thrift OpenSSL support by @egolearner in #819

New Contributors

Full Changelog: v0.7.0...v0.8.0-alpha

Don't miss a new zvec release

NewReleases is sending notifications on new releases.