[NIGHTLY] v24.12.00

🔗 Links

• Development Branch
• Compare with main branch

🛠️ Improvements

• Merge branch-24.10 into branch-24.12 (#2461) @jameslamb

v24.08.00

🚨 Breaking Changes

• [Refactor] move popc to under util (#2394) @rhdong
• [Opt] Expose the detail::popc as public API (#2346) @rhdong

🐛 Bug Fixes

• Add timeout to UCXX generic operations (#2398) @pentschev
• [Fix] bitmap set/test issue (#2371) @rhdong
• Fix 0 recall issue in raft_cagra_hnswlib ANN benchmark (#2369) @divyegala
• Fix ef setting in HNSW wrapper (#2367) @divyegala
• Fix cagra graph opt bug (#2365) @enp1s0
• Fix a bug where the wrong API is used to free the memory (#2361) @PointKernel
• Allow anonymous user in devcontainer name (#2355) @bdice
• Fix compilation error when _CLK_BREAKDOWN is defined in cagra. (#2350) @jiangyinzuo
• ensure raft-dask wheel tests install pylibraft wheel from the same CI run, fix wheel dependencies (#2349) @jameslamb
• Change --config-setting to --config-settings (#2342) @KyleFromNVIDIA
• Add workaround for syevd in CUDA 12.0 (#2332) @lowener

🚀 New Features

• [FEA] add the support of masked_matmul (#2362) @rhdong
• [FEA] Dice Distance for Dense Inputs (#2359) @aamijar
• [Opt] Expose the detail::popc as public API (#2346) @rhdong
• Enable distance return for NN Descent (#2345) @jinsolp

🛠️ Improvements

• [Refactor] move popc to under util (#2394) @rhdong
• split up CUDA-suffixed dependencies in dependencies.yaml (#2388) @jameslamb
• Use workflow branch 24.08 again (#2385) @KyleFromNVIDIA
• Add cusparseSpMV_preprocess to cusparse wrapper (#2384) @Kh4ster
• Consolidate SUM reductions (#2381) @mfoerste4
• Use slicing kernel to copy distances inside NN Descent (#2380) @jinsolp
• Build and test with CUDA 12.5.1 (#2378) @KyleFromNVIDIA
• Add CUDA_STATIC_MATH_LIBRARIES (#2376) @KyleFromNVIDIA
• skip CMake 3.30.0 (#2375) @jameslamb
• Use verify-alpha-spec hook (#2373) @KyleFromNVIDIA
• Binarize Dice Distance for Dense Inputs (#2370) @aamijar
• [FEA] Add distance epilogue for NN Descent (#2364) @jinsolp
• resolve dependency-file-generator warning, other rapids-build-backend followup (#2360) @jameslamb
• Remove text builds of documentation (#2354) @vyasr
• Use default init in reduction (#2351) @akifcorduk
• ensure update-version.sh preserves alpha spec, add tests on version constants (#2344) @jameslamb
• remove unnecessary 'setuptools' dependencies (#2343) @jameslamb
• Use rapids-build-backend (#2331) @KyleFromNVIDIA
• Add FAISS with RAFT enabled Benchmarking to raft-ann-bench (#2026) @tarang-jain

[NIGHTLY] v24.10.00

🔗 Links

• Development Branch
• Compare with main branch

🚨 Breaking Changes

• [Feat] add repeat, sparsity, eval_n_elements APIs to bitset (#2439) @rhdong

🐛 Bug Fixes

• Disable NN Descent Batch tests temporarily (#2453) @divyegala
• Fix sed syntax in update-version.sh (#2441) @raydouglass
• Use runtime check of cudart version for eig (#2430) @lowener
• [BUG] Fix bitset function visibility (#2429) @lowener
• Exclude any kernel symbol that uses cutlass (#2425) @robertmaynard

🚀 New Features

• [Feat] add repeat, sparsity, eval_n_elements APIs to bitset (#2439) @rhdong
• [Opt] Enforce the UT Coverity and add benchmark for transpose (#2438) @rhdong
• [FEA] Support for half-float mixed precise in brute-force (#2382) @rhdong

🛠️ Improvements

• bump NCCL floor to 2.19 (#2458) @jameslamb
• Deprecating vector search APIs and updating README accordingly (#2448) @cjnolet
• Update update-version.sh to use packaging lib (#2447) @AyodeAwe
• Switch traceback to native (#2446) @galipremsagar
• bump NCCL floor to 2.18.1.1 (#2443) @jameslamb
• Add missing cuda_suffixed: true (#2440) @trxcllnt
• Use CI workflow branch 'branch-24.10' again (#2437) @jameslamb
• Update to flake8 7.1.1. (#2435) @bdice
• Update fmt (to 11.0.2) and spdlog (to 1.14.1). (#2433) @jameslamb
• Allow coo_sort to work on int64_t indices (#2432) @benfred
• Adding NCCL clique to the RAFT handle (#2431) @viclafargue
• Add support for Python 3.12 (#2428) @jameslamb
• Update rapidsai/pre-commit-hooks (#2420) @KyleFromNVIDIA
• Drop Python 3.9 support (#2417) @jameslamb
• Use CUDA math wheels (#2415) @KyleFromNVIDIA
• Remove NumPy <2 pin (#2414) @seberg
• Update pre-commit hooks (#2409) @KyleFromNVIDIA
• Improve update-version.sh (#2408) @bdice
• Use tool.scikit-build.cmake.version, set scikit-build-core minimum-version (#2406) @jameslamb
• [FEA] Batching NN Descent (#2403) @jinsolp
• Update pip devcontainers to UCX v1.17.0 (#2401) @jameslamb
• Merge branch-24.08 into branch-24.10 (#2397) @jameslamb

v24.06.00

🚨 Breaking Changes

• Rename raft-ann-bench module to raft_ann_bench (#2333) @KyleFromNVIDIA
• Scaling workspace resources (#2322) @achirkin
• [REVIEW] Adjust UCX dependencies (#2304) @pentschev
• Convert device_memory_resource* to device_async_resource_ref (#2269) @harrism

🐛 Bug Fixes

• Fix import of VERSION file in raft-ann-bench (#2338) @KyleFromNVIDIA
• Rename raft-ann-bench module to raft_ann_bench (#2333) @KyleFromNVIDIA
• Support building faiss main statically (#2323) @robertmaynard
• Refactor spectral scale_obs to use existing normalization function (#2319) @ChuckHastings
• Correct initializer list order found by cuvs (#2317) @robertmaynard
• ANN_BENCH: enable move semantics for configured_raft_resources (#2311) @achirkin
• Revert "Build C++ wheel (#2264)" (#2305) @vyasr
• Revert "Add compile-library by default on pylibraft build" (#2300) @vyasr
• Add VERSION to raft-ann-bench package (#2299) @KyleFromNVIDIA
• Remove nonexistent job from workflow (#2298) @vyasr
• libucx should be run dependency of raft-dask (#2296) @divyegala
• Fix clang intrinsic warning (#2292) @aaronmondal
• Replace too long index file name with hash in ANN bench (#2280) @tfeher
• Fix build command for C++ compilation (#2270) @lowener
• Fix a compilation error in CAGRA when enabling log output (#2262) @enp1s0
• Correct member initialization order (#2254) @robertmaynard
• Fix time computation in CAGRA notebook (#2231) @lowener

📖 Documentation

• Fix citation info (#2318) @enp1s0

🚀 New Features

• Scaling workspace resources (#2322) @achirkin
• ANN_BENCH: AnnGPU::uses_stream() for optional algo GPU sync (#2314) @achirkin
• [FEA] Split Bitset code (#2295) @lowener
• [FEA] support of prefiltered brute force (#2294) @rhdong
• Always use a static gtest and gbench (#2265) @robertmaynard
• Build C++ wheel (#2264) @vyasr
• InnerProduct Distance Metric for CAGRA search (#2260) @tarang-jain
• [FEA] Add support for select_k on CSR matrix (#2140) @rhdong

🛠️ Improvements

• ANN_BENCH: common AnnBase::index_type (#2315) @achirkin
• ANN_BENCH: split instances of RaftCagra into multiple files (#2313) @achirkin
• ANN_BENCH: a global pool of result buffers across benchmark cases (#2312) @achirkin
• Remove the shared state and the mutex from NVTX internals (#2310) @achirkin
• docs: update README.md (#2308) @eltociear
• [REVIEW] Reenable raft-dask wheel tests requiring UCX-Py (#2307) @pentschev
• [REVIEW] Adjust UCX dependencies (#2304) @pentschev
• Overhaul ops-codeowners (#2303) @raydouglass
• Make thrust nosync execution policy the default thrust policy (#2302) @abc99lr
• InnerProduct testing for CAGRA+HNSW (#2297) @divyegala
• Enable warnings as errors for Python tests (#2288) @mroeschke
• Normalize dataset vectors in the CAGRA InnerProduct tests (#2287) @enp1s0
• Use dynamic version for raft-ann-bench (#2285) @KyleFromNVIDIA
• Make 'librmm' a 'host' dependency for conda packages (#2284) @jameslamb
• Fix comments in cpp/include/raft/neighbors/cagra_serialize.cuh (#2283) @jiangyinzuo
• Only use functions in the limited API (#2282) @vyasr
• define 'ucx' pytest marker (#2281) @jameslamb
• Migrate to {{ stdlib("c") }} (#2278) @hcho3
• add --rm and --name to devcontainer run args (#2275) @trxcllnt
• Update pip devcontainers to UCX v1.15.0 (#2274) @trxcllnt
• #ifdef out pragma deprecation warning messages (#2271) @trxcllnt
• Convert device_memory_resource* to device_async_resource_ref (#2269) @harrism
• Update the developer's guide with new copyright hook (#2266) @KyleFromNVIDIA
• Improve coalesced reduction performance for tall and thin matrices (up to 2.6x faster) (#2259) @Nyrio
• Adds missing files to update-version.sh (#2255) @AyodeAwe
• Enable all tests for arm64 jobs (#2248) @galipremsagar
• Update nvtx3 link in cmake (#2246) @lowener
• Add CAGRA-Q subspace dim = 4 support (#2244) @enp1s0
• Get rid of cuco::sentinel namespace (#2243) @PointKernel
• Replace usages of raw get_upstream with get_upstream_resource() (#2207) @miscco
• Set the import mode for dask tests (#2142) @vyasr
• Add UCXX support (#1983) @pentschev

v24.04.00

🐛 Bug Fixes

• Update pre-commit-hooks to v0.0.3 (#2239) @KyleFromNVIDIA
• MAINT: Simplify NCCL worker rank identification (#2228) @VibhuJawa
• Fix bug in blockRankedReduce (#2226) @akifcorduk
• Fix illegal acces mean/stdev, sum add Kahan Summation (#2223) @mfoerste4
• Batch cutlass distance kernels along N matrix dim (#2215) @mdoijade
• Fix out of bounds access in sum kernel (#2183) @tfeher
• Fix ANN bench ground truth generation for k>1024 (#2180) @tfeher
• Fixing cusparse aligned address issue and adding note (#2179) @cjnolet
• Launch neighborhood_recall kernel on CUDA stream (#2156) @divyegala
• Add compile-library by default on pylibraft build (#2090) @lowener

📖 Documentation

• Adding cuVS notice to README and front page of docs. (#2224) @cjnolet

🚀 New Features

• Add CAGRA-Q to ANN benchmarks (#2233) @achirkin
• Add CAGRA-Q build (compression) (#2213) @achirkin
• CAGRA-Q search (#2206) @enp1s0
• Demangle backtrace symbols on raft error (#2188) @achirkin
• Reapply: Support for fp16 in CAGRA and IVF-PQ (#2172) @achirkin
• Remove supports_streams from custom RAFT memory resources (#2121) @harrism
• [FEA] Add support for bitmap_view & the API of bitmap_to_csr (#2109) @rhdong

🛠️ Improvements

• Use conda env create --yes instead of --force (#2247) @bdice
• Align ucx version pinning with ucx-py/ucxx. (#2227) @bdice
• Add upper bound to prevent usage of NumPy 2 (#2222) @bdice
• Performance optimization of IVF-flat / select_k (#2221) @mfoerste4
• Replace local copyright check with pre-commit-hooks verify-copyright (#2220) @KyleFromNVIDIA
• Remove hard-coding of RAPIDS version where possible (#2219) @KyleFromNVIDIA
• Fix style. (#2214) @bdice
• Add explicit instantiations for IVF-PQ search kernels used in tests (#2212) @tfeher
• Improve RBC eps-neighborhood query performance (#2211) @mfoerste4
• Add test for spmm (#2210) @mfoerste4
• Only install necessary components in conda packages. (#2209) @bdice
• Automate C++ include file grouping and ordering using clang-format (#2202) @harrism
• Add support for Python 3.11, require NumPy 1.23+ (#2200) @jameslamb
• Pass std::optional instead of thrust::optional to RMM (#2199) @trxcllnt
• Update devcontainers to CUDA Toolkit 12.2 (#2192) @trxcllnt
• target branch-24.04 for GitHub Actions workflows (#2189) @jameslamb
• Fixing workaround for cuSPARSE bug with correct copy dimensions (#2185) @mfoerste4
• Allow topk larger than 1024 in CAGRA (#2181) @benfred
• IVF-FLAT support k > 256 (#2169) @mfoerste4
• Add environment-agnostic scripts for running ctests and pytests (#2165) @trxcllnt
• Ensure that ctest is called with --no-tests=error. (#2163) @bdice
• Update ops-bot.yaml (#2158) @AyodeAwe
• random sampling of dataset rows with improved memory utilization (#2155) @tfeher
• [FIX] Ensure hnswlib can be found from RAFT's build dir (#2145) @trxcllnt
• Improve analysis experience for ANN benchmarks (#2139) @achirkin
• Enable CAGRA index building without adding dataset to the index (#2126) @tfeher
• Add fused cosine 1-NN cutlass based kernel (#2125) @mdoijade
• Update raft for compatibility with the latest cuco (#2118) @PointKernel
• Support CUDA 12.2 (#2092) @jameslamb
• Cache IVF-PQ and select-warpsort kernel launch parameters to reduce latency (#1786) @achirkin

v24.02.00

🚨 Breaking Changes

• Switch to scikit-build-core (#2051) @vyasr
• Update to CCCL 2.2.0. (#2049) @bdice
• Update raft-ann-bench output filenames and add features to plotting (#2043) @divyegala
• Remove selection_faiss (#2027) @benfred

🐛 Bug Fixes

• fix is_row/col_order for strided layouts (#2173) @mfoerste4
• Fix failing C++ tests and revert #2097, #2085. (#2168) @cjnolet
• Exclude tests from builds (#2162) @vyasr
• [HOTFIX] 24.02 Revert Random Sampling (#2144) @cjnolet
• Pin to pytest 7. (#2137) @bdice
• Conditionally include hnsw wrapper source in CMake (#2135) @divyegala
• [BUG] Fix SPMM strided view (#2124) @lowener
• Fixing small bug in CUSPARSE spmm w/ CUDA 12.2 (#2117) @cjnolet
• [BUG] Fix num_cta_per_query div (#2107) @lowener
• Remove extraneous host pinnings from libraft-headers-only. (#2102) @bdice
• Remove unneeded CI symbol excludes (#2098) @robertmaynard
• Properly taking ownership of nccl subcomm (and destroying it) (#2094) @cjnolet
• Fix max_queries for CAGRA (#2081) @lowener
• Fix compile failure on RTX 4090 (#2076) @JieFengWang
• Fix a crash in FAISS benchmark wrapper introduced in #2021 (#2062) @achirkin
• Correct function that wasn't returning a value (#2045) @robertmaynard
• Fixing small bug in raft-ann-bench (#2041) @cjnolet
• Make device_resources accessed from device_resources_manager thread-safe (#2030) @wphicks
• Fix ann-bench multithreading (#2021) @achirkin
• Fix ci/checks/copyright.py to mirror RAPIDS reference (#2008) @divyegala
• Fix pyproject versions (#2002) @vyasr

📖 Documentation

• Adding license info for wiki-all dataset (#2129) @cjnolet
• [DOC] Documentation updates for release 24.02 (#2093) @cjnolet
• Fix errors with ingroup exposed by doxygen 1.10 (#2079) @wphicks
• Fix a typo (#2070) @narangvivek10
• Add usage example for brute_force::build (#2029) @benfred
• Add filtering to vector search tutorial (#1996) @lowener

🚀 New Features

• Update to use rapids-cmake for all deps (#2096) @robertmaynard
• Add IVF-PQ example into the template project (#2091) @achirkin
• Support for fp16 in CAGRA and IVF-PQ (#2085) @achirkin
• Add random subsampling for IVF methods (#2077) @tfeher
• Update raft-ann-bench output filenames and add features to plotting (#2043) @divyegala
• Add brute_force index serialization (#2036) @wphicks
• Add eps-neighbor search via RBC (#2028) @mfoerste4
• libraft and pylibraft API for CAGRA build and HNSW search (#2022) @divyegala
• Export Pareto frontier in raft-ann-bench.data_export (#2009) @divyegala
• Implement maybe-owning multi-dimensional container (mdbuffer) (#1999) @wphicks
• Add support for 1024+ dim vectors in CAGRA search (#1994) @enp1s0
• Replace GEMM backend: cublas.gemm -> cublaslt.matmul (#1736) @achirkin

🛠️ Improvements

• Remove get_mem_info functions from RAFT custom memory resources (#2108) @harrism
• Replace call to mr::get_mem_info() (#2099) @harrism
• Allow topk larger than 1024 in CAGRA (#2097) @benfred
• Remove usages of rapids-env-update (#2095) @KyleFromNVIDIA
• Provide explicit pool size for pool_memory_resources and clean up includes (#2088) @harrism
• refactor CUDA versions in dependencies.yaml (#2086) @jameslamb
• ANN bench fix latency measurement overhead (#2084) @tfeher
• Remove hardcoded limit in print_results function (#2080) @narangvivek10
• [FEA] Add support for SDDMM by wrapping the cusparseSDDMM (#2067) (#2067) @rhdong
• Benchmark brute force knn (#2063) @benfred
• [BUG] fix empty initialization of device_ndarray in pylibraft (#2061) @mfoerste4
• Improve parallelism of refine host (#2059) @anaruse
• Subsampling for IVF-PQ codebook generation (#2052) @abc99lr
• Switch to scikit-build-core (#2051) @vyasr
• Update to CCCL 2.2.0. (#2049) @bdice
• Use cuda::proclaim_return_type on device lambda. (#2048) @bdice
• Removing code that explicitly compares equality of rmm memory resources (#2047) @cjnolet
• Add public enum for select-k algorithm selection (#2046) @benfred
• Update dependencies.yaml to new pip index (#2042) @vyasr
• Remove RAFT_BUILD_WHEELS and standardize Python builds (#2040) @vyasr
• Fix ucx-py version pinning in dependencies.yaml. (#2035) @bdice
• [REVIEW] Fix typos in parameter tuning guide (#2034) @abc99lr
• Add AIR-Top-k reference (#2031) @tfeher
• Remove selection_faiss (#2027) @benfred
• Fixing json parse error in raft-ann-bench.data_export (#2025) @cjnolet
• Updating cagra build constraint (#2016) @cjnolet
• Update to fmt 10.1.1 and spdlog 1.12.0. (#1957) @bdice
• Enable host dataset for IVF-Flat (#1635) @tfeher
• add half/bfloat support to myInf and abs (#1592) @Kh4ster

v23.12.00

🐛 Bug Fixes

• Update actions/labeler to v4 (#2037) @raydouglass
• pylibraft only depends on numpy at runtime, not build time. (#2013) @bdice
• Fixes to update-version.sh (#1991) @raydouglass
• Adjusting end-to-end start time so it doesn't include stream creation time (#1989) @cjnolet
• CAGRA graph optimizer: clamp rev_graph_count (#1987) @tfeher
• Catching conversion errors in data_export instead of fully failing (#1979) @cjnolet
• Fix syncing mechanism in raft-ann-bench C++ search (#1961) @divyegala
• Fixing hnswlib in latency mode (#1959) @cjnolet
• Fix ucx-py alpha version update for raft-dask (#1953) @divyegala
• Reduce NN Descent test threshold (#1946) @divyegala
• Fixes to new YAML config raft-bench-ann (#1945) @divyegala
• Set RNG seeds in NN Descent to diagnose flaky tests (#1931) @divyegala
• Fix FAISS CPU algorithm names in raft-ann-bench (#1916) @divyegala
• Increase iterations in NN Descent tests to avoid flakiness (#1915) @divyegala
• Fix filepath in raft-ann-bench/split_groundtruth module (#1911) @divyegala
• Remove dynamic entry-points from raft-ann-bench (#1910) @benfred
• Remove unnecessary dataset path check in ANN bench (#1908) @tfeher
• Fixing Googletests and re-enabling in CI (#1904) @cjnolet
• Fix NN Descent overflows (#1875) @divyegala
• Build fix for CUDA 12.2 (#1870) @benfred
• [BUG] Fix a bug in NN descent (#1869) @enp1s0

📖 Documentation

• Brute Force Index documentation fix (#1944) @lowener
• Add wiki_all dataset config and documentation. (#1918) @cjnolet
• Updates to raft-ann-bench docs (#1905) @cjnolet
• End-to-end vector search tutorial in docs (#1776) @cjnolet

🚀 New Features

• Adding dry-run option to raft-ann-bench (#1970) @cjnolet
• Add ANN bench scripts to generate ground truth (#1967) @tfeher
• CAGRA build + HNSW search (#1956) @divyegala
• Verify conda-cpp-post-build-checks (#1935) @robertmaynard
• Make all cuda kernels have hidden visibility (#1898) @robertmaynard
• Update rapids-cmake functions to non-deprecated signatures (#1884) @robertmaynard
• [FEA] Helpers for identifying contiguous layouts. (#1861) @trivialfis
• Add raft::stats::neighborhood_recall (#1860) @divyegala
• [FEA] Helpers and CodePacker for IVF-PQ (#1826) @tarang-jain

🛠️ Improvements

• Pinning fmt and spdlog for raft-ann-bench-cpu (#2018) @cjnolet
• Build concurrency for nightly and merge triggers (#2011) @bdice
• Using EXPORT_SET in rapids_find_package_root (#2006) @cjnolet
• Remove static checks for serialization size (#1997) @cjnolet
• Skipping bad json parse (#1990) @cjnolet
• Update select-k heuristic (#1985) @benfred
• ANN bench: use different offset for each thread (#1981) @tfeher
• Allow raft-ann-bench/run to continue after encountering bad YAML configs (#1980) @divyegala
• Add build and search params to raft-ann-bench.data_export CSVs (#1971) @divyegala
• Use new rapids-dask-dependency metapackage for managing dask versions (#1968) @galipremsagar
• Remove unused header (#1960) @wphicks
• Adding pool back in and fixing cagra benchmark params (#1951) @cjnolet
• Add constraints to hnswlib in raft-bench-ann (#1949) @divyegala
• Add support for iterating over batches in bfknn (#1947) @benfred
• Fix ANN bench latency (#1940) @tfeher
• Add YAML config files to run parameter sweeps for ANN benchmarks (#1929) @divyegala
• Relax ucx pinning (#1927) @vyasr
• Try using contiguous rank to fix cuda_visible_devices (#1926) @VibhuJawa
• Unpin dask and distributed for 23.12 development (#1925) @galipremsagar
• Adding throughput and latency modes to raft-ann-bench (#1920) @cjnolet
• Providing aarch64 yaml environment files (#1914) @cjnolet
• CAGRA ANN bench: parse build options for IVF-PQ build algo (#1912) @tfeher
• Fix python script location in ANN bench description (#1906) @tfeher
• Refactor install/build guide. (#1899) @cjnolet
• Check return values of raft-ann-bench subprocess calls (#1897) @benfred
• ANN bench options to specify CAGRA graph and dataset locations (#1896) @cjnolet
• Add check-json to pre-commit linters, and fix invalid ann-bench JSON config (#1894) @benfred
• Use branch-23.12 workflows. (#1886) @bdice
• Setup Consistent Nightly Versions for Pip and Conda (#1880) @divyegala
• Fix and improve one-block radix select (#1878) @yong-wang
• [FEA] Improvements on bitset class (#1877) @lowener
• Branch 23.12 merge 23.10 (#1873) @AyodeAwe
• Branch 23.12 merge 23.10 (#1868) @cjnolet
• Replace raft::random calls to not use deprecated API (#1867) @lowener
• raft: Build CUDA 12.0 ARM conda packages. (#1853) @bdice
• Documentation for raft ANN benchmark containers. (#1833) @dantegd
• [FEA] Support vector deletion in ANN IVF (#1831) @lowener
• Provide a raft::copy overload for mdspan-to-mdspan copies (#1818) @wphicks
• Adding FAISS cpu to raft-ann-bench (#1814) @cjnolet

v23.10.00

🚨 Breaking Changes

• Change CAGRA auto mode selection (#1830) @enp1s0
• Update CAGRA serialization (#1755) @benfred
• Improvements to ANN Benchmark Python scripts and docs (#1734) @divyegala
• Update to Cython 3.0.0 (#1688) @vyasr
• ANN-benchmarks: switch to use gbench (#1661) @achirkin

🐛 Bug Fixes

• [BUG] Fix a bug in the filtering operation in CAGRA multi-kernel (#1862) @enp1s0
• Fix conf file for benchmarking glove datasets (#1846) @dantegd
• raft-ann-bench package fixes for plotting and conf files (#1844) @dantegd
• Fix update-version.sh for all pyproject.toml files (#1839) @raydouglass
• Make RMM a run dependency of the raft-ann-bench conda package (#1838) @dantegd
• Printing actual exception in require base set (#1816) @cjnolet
• Adding rmm to raft-ann-bench dependencies (#1815) @cjnolet
• Use conda mambabuild not mamba mambabuild (#1812) @bdice
• Fix raft-dask naming in wheel builds (#1805) @divyegala
• neighbors::refine_host: check the dataset bounds (#1793) @achirkin
• [BUG] Fix search parameter check in CAGRA (#1784) @enp1s0
• IVF-Flat: fix search batching (#1764) @achirkin
• Using expanded distance computations in pylibraft (#1759) @cjnolet
• Fix ann-bench Documentation (#1754) @divyegala
• Make get_cache_idx a weak symbol with dummy template (#1733) @ahendriksen
• Fix IVF-PQ fused kernel performance problems (#1726) @achirkin
• Fix build.sh to enable NEIGHBORS_ANN_CAGRA_TEST (#1724) @enp1s0
• Fix template types for create_descriptor function. (#1680) @csadorf

📖 Documentation

• Fix the CAGRA paper citation (#1788) @enp1s0
• Add citation info for the CAGRA paper preprint (#1787) @enp1s0
• [DOC] Fix grouping for ANN in C++ doxygen (#1782) @lowener
• Update RAFT documentation (#1717) @lowener
• Additional polishing of README and docs (#1713) @cjnolet

🚀 New Features

• [FEA] Add bitset_filter for CAGRA indices removal (#1837) @lowener
• ann-bench: miscellaneous improvements (#1808) @achirkin
• [FEA] Add bitset for ANN pre-filtering and deletion (#1803) @lowener
• Adding config files for remaining (relevant) ann-benchmarks million-scale datasets (#1761) @cjnolet
• Port NN-descent algorithm to use in cagra::build() (#1748) @divyegala
• Adding conda build for libraft static (#1746) @cjnolet
• [FEA] Provide device_resources_manager for easy generation of device_resources (#1716) @wphicks

🛠️ Improvements

• Add option to brute_force index to store non-owning reference to norms (#1865) @benfred
• Pin dask and distributed for 23.10 release (#1864) @galipremsagar
• Update image names (#1835) @AyodeAwe
• Fixes for OOM during CAGRA benchmarks (#1832) @benfred
• Change CAGRA auto mode selection (#1830) @enp1s0
• Update to clang 16.0.6. (#1829) @bdice
• Add IVF-Flat C++ example (#1828) @tfeher
• matrix::select_k: extra tests and benchmarks (#1821) @achirkin
• Add index class for brute_force knn (#1817) @benfred
• [FEA] Add pre-filtering to CAGRA (#1811) @enp1s0
• More updates to ann-bench docs (#1810) @cjnolet
• Add best deep-100M configs for IVF-PQ to ANN benchmarks (#1807) @tfeher
• A few fixes to raft-ann-bench recipe and docs (#1806) @cjnolet
• Simplify wheel build scripts and allow alphas of RAPIDS dependencies (#1804) @divyegala
• Various fixes to reproducible benchmarks (#1800) @cjnolet
• ANN-bench: more flexible cuda_stub.hpp (#1792) @achirkin
• Add RAFT devcontainers (#1791) @trxcllnt
• Cagra memory optimizations (#1790) @benfred
• Fixing a couple security concerns in raft-dask nccl unique id generation (#1785) @cjnolet
• Don't serialize dataset with CAGRA bench (#1781) @benfred
• Use copy-pr-bot (#1774) @ajschmidt8
• Add GPU and CPU packages for ANN benchmarks (#1773) @dantegd
• Improvements to raft-ann-bench scripts, docs, and benchmarking implementations. (#1769) @cjnolet
• [REVIEW] Introducing host API for PCG (#1767) @vinaydes
• Unpin dask and distributed for 23.10 development (#1760) @galipremsagar
• Add ivf-flat notebook (#1758) @tfeher
• Update CAGRA serialization (#1755) @benfred
• Remove block size template parameter from CAGRA search (#1740) @enp1s0
• Add NVTX ranges for cagra search/serialize functions (#1737) @benfred
• Improvements to ANN Benchmark Python scripts and docs (#1734) @divyegala
• Fixing forward merger for 23.08 -> 23.10 (#1731) @cjnolet
• [FEA] Use CAGRA in C++ template (#1730) @lowener
• fixed box around raft image (#1710) @nwstephens
• Enable CUTLASS-based distance kernels on CTK 12 (#1702) @ahendriksen
• Update bench-ann configuration (#1696) @lowener
• Update to Cython 3.0.0 (#1688) @vyasr
• Update CMake version (#1677) @vyasr
• Branch 23.10 merge 23.08 (#1672) @vyasr
• ANN-benchmarks: switch to use gbench (#1661) @achirkin

v23.08.00

🚨 Breaking Changes

• Separate CAGRA index type from internal idx type (#1664) @tfeher
• Stop using setup.py in build.sh (#1645) @vyasr
• CAGRA max_queries auto configuration (#1613) @enp1s0
• Rename the CAGRA prune function to optimize (#1588) @enp1s0
• CAGRA pad dataset for 128bit vectorized load (#1505) @tfeher
• Sparse Pairwise Distances API Updates (#1502) @divyegala
• Cagra index construction without copying device mdarrays (#1494) @tfeher
• [FEA] Masked NN for connect_components (#1445) @tarang-jain
• Limiting workspace memory resource (#1356) @achirkin

🐛 Bug Fixes

• Remove push condition on docs-build (#1693) @raydouglass
• IVF-PQ: Fix illegal memory access with large max_samples (#1685) @achirkin
• Fix missing parameter for select_k (#1682) @ucassjy
• Separate CAGRA index type from internal idx type (#1664) @tfeher
• Add rmm to pylibraft run dependencies, since it is used by Cython. (#1656) @bdice
• Hotfix: wrong constant in IVF-PQ fp_8bit2half (#1654) @achirkin
• Fix sparse KNN for large batches (#1640) @viclafargue
• Fix uploading of RAFT nightly packages (#1638) @dantegd
• Fix cagra multi CTA bug (#1628) @enp1s0
• pass correct stream to cutlass kernel launch of L2/cosine pairwise distance kernels (#1597) @mdoijade
• Fix launchconfig y-gridsize too large in epilogue kernel (#1586) @mfoerste4
• Fix update version and pinnings for 23.08. (#1556) @bdice
• Fix for function exposing KNN merge (#1418) @viclafargue

📖 Documentation

• Critical doc fixes and updates for 23.08 (#1705) @cjnolet
• Fix the documentation about changing the logging level (#1596) @enp1s0
• Fix raft::bitonic_sort small usage example (#1580) @enp1s0

🚀 New Features

• Use rapids-cmake new parallel testing feature (#1623) @robertmaynard
• Add support for row-major slice (#1591) @lowener
• IVF-PQ tutorial notebook (#1544) @achirkin
• [FEA] Masked NN for connect_components (#1445) @tarang-jain
• raft: Build CUDA 12 packages (#1388) @vyasr
• Limiting workspace memory resource (#1356) @achirkin

🛠️ Improvements

• Pin dask and distributed for 23.08 release (#1711) @galipremsagar
• Add algo parameter for CAGRA ANN bench (#1687) @tfeher
• ANN benchmarks python wrapper for splitting billion-scale dataset groundtruth (#1679) @divyegala
• Rename CAGRA parameter num_parents to search_width (#1676) @tfeher
• Renaming namespaces to promote CAGRA from experimental (#1666) @cjnolet
• CAGRA Python wrappers (#1665) @dantegd
• Add notebook for Vector Search - Question Retrieval (#1662) @lowener
• Fix CMake CUDA support for pylibraft when raft is found. (#1659) @bdice
• Cagra ANN benchmark improvements (#1658) @tfeher
• ANN-benchmarks: avoid using the dataset during search when possible (#1657) @achirkin
• Revert CUDA 12.0 CI workflows to branch-23.08. (#1652) @bdice
• ANN: Optimize host-side refine (#1651) @achirkin
• Cagra template instantiations (#1650) @tfeher
• Modify comm_split to avoid ucp (#1649) @ChuckHastings
• Stop using setup.py in build.sh (#1645) @vyasr
• IVF-PQ: Add a (faster) direct conversion fp8->half (#1644) @achirkin
• Simplify bench/ann scripts to Python based module (#1642) @divyegala
• Further removal of uses-setup-env-vars (#1639) @dantegd
• Drop blank line in raft-dask/meta.yaml (#1637) @jakirkham
• Enable conservative memory allocations for RAFT IVF-Flat benchmarks. (#1634) @tfeher
• [FEA] Codepacking for IVF-flat (#1632) @tarang-jain
• Fixing ann bench cmake (and docs) (#1630) @cjnolet
• [WIP] Test CI issues (#1626) @VibhuJawa
• Set pool memory resource for raft IVF ANN benchmarks (#1625) @tfeher
• Adding sort option to matrix::select_k api (#1615) @cjnolet
• CAGRA max_queries auto configuration (#1613) @enp1s0
• Use exceptions instead of exit(-1) (#1594) @benfred
• [REVIEW] Add scheduler_file argument to support MNMG setup (#1593) @VibhuJawa
• Rename the CAGRA prune function to optimize (#1588) @enp1s0
• This PR adds support to __half and nb_bfloat16 to myAtomicReduce (#1585) @Kh4ster
• [IMP] move core CUDA RT macros to cuda_rt_essentials.hpp (#1584) @MatthiasKohl
• preprocessor syntax fix (#1582) @AyodeAwe
• use rapids-upload-docs script (#1578) @AyodeAwe
• Unpin dask and distributed for development and fix merge_labels test (#1574) @galipremsagar
• Remove documentation build scripts for Jenkins (#1570) @ajschmidt8
• Add support to __half and nv_bfloat16 to most math functions (#1554) @Kh4ster
• Add RAFT ANN benchmark for CAGRA (#1552) @enp1s0
• Update CAGRA knn_graph_sort to use Raft::bitonic_sort (#1550) @enp1s0
• Add identity matrix function (#1548) @lowener
• Unpin scikit-build upper bound (#1547) @vyasr
• Migrate wheel workflow scripts locally (#1546) @divyegala
• Add sample filtering for ivf_flat. Filtering code refactoring and cleanup (#1541) @alexanderguzhva
• CAGRA pad dataset for 128bit vectorized load (#1505) @tfeher
• Sparse Pairwise Distances API Updates (#1502) @divyegala
• Add CAGRA gbench (#1496) @tfeher
• Cagra index construction without copying device mdarrays (#1494) @tfeher

v23.06.02

🚨 Breaking Changes

• ivf-pq::search: fix the indexing type of the query-related mdspan arguments (#1539) @achirkin
• Dropping Python 3.8 (#1454) @divyegala

🐛 Bug Fixes

• PR to trigger rebuild of raft (#1622) @raydouglass
• Assigning Deterministic rank to Dask Workers Based on CUDA_VISIBLE_DEVICES (branch-23.06) (#1587) @VibhuJawa
• [HOTFIX] Fix distance metrics L2/cosine/correlation when X & Y are same buffer but with different shape and add unit test for such case. (#1571) @mdoijade
• Using raft::resources in rsvd (#1543) @cjnolet
• ivf-pq::search: fix the indexing type of the query-related mdspan arguments (#1539) @achirkin
• Check python brute-force knn inputs (#1537) @benfred
• Fix failing TiledKNNTest unittest (#1533) @benfred
• ivf-flat: fix incorrect recomputed size of the index (#1525) @achirkin
• ivf-flat: limit the workspace size of the search via batching (#1515) @achirkin
• Support uint64_t in CAGRA index data type (#1514) @enp1s0
• Workaround for cuda 12 issue in cusparse (#1508) @cjnolet
• Un-scale output distances (#1499) @achirkin
• Inline get_cache_idx (#1492) @ahendriksen
• Pin to scikit-build<17.2 (#1487) @vyasr
• Remove pool_size() calls from debug printouts (#1484) @tfeher
• Add missing ext declaration for log detail::format (#1482) @tfeher
• Remove include statements from inside namespace (#1467) @robertmaynard
• Use pin_compatible to ensure that lower CTKs can be used (#1462) @vyasr
• fix ivf_pq n_probes (#1456) @benfred
• The glog project root CMakeLists.txt is where we should build from (#1442) @robertmaynard
• Add missing resource factory virtual destructor (#1433) @cjnolet
• Removing cuda stream view include from mdarray (#1429) @cjnolet
• Fix dim param for IVF-PQ wrapper in ANN bench (#1427) @tfeher
• Remove MetricProcessor code from brute_force::knn (#1426) @benfred
• Fix is_min_close (#1419) @benfred
• Have consistent compile lines between BUILD_TESTS enabled or not (#1401) @robertmaynard
• Fix ucx-py pin in raft-dask recipe (#1396) @vyasr

📖 Documentation

• Various updates to the docs for 23.06 release (#1538) @cjnolet
• Rename kernel arch finding function for dispatch (#1536) @mdoijade
• Adding bfknn and ivf-pq python api to docs (#1507) @cjnolet
• Add RAPIDS cuDF as a library that supports cuda_array_interface (#1444) @miguelusque

🚀 New Features

• IVF-PQ: manipulating individual lists (#1298) @achirkin
• Gram matrix support for sparse input (#1296) @mfoerste4
• [FEA] Add randomized svd from cusolver (#1000) @lowener

🛠️ Improvements

• Require Numba 0.57.0+ (#1559) @jakirkham
• remove device_resources include from linalg::map (#1540) @benfred
• Learn heuristic to pick fastest select_k algorithm (#1523) @benfred
• [REVIEW] make raft::cache::Cache protected to allow overrides (#1522) @mfoerste4
• [REVIEW] Fix padding assertion in sparse Gram evaluation (#1521) @mfoerste4
• run docs nightly too (#1520) @AyodeAwe
• Switch back to using primary shared-action-workflows branch (#1519) @vyasr
• Python API for IVF-Flat serialization (#1516) @tfeher
• Introduce sample filtering to IVFPQ index search (#1513) @alexanderguzhva
• Migrate from raft::device_resources -> raft::resources (#1510) @benfred
• Use rmm allocator in CAGRA prune (#1503) @enp1s0
• Update recipes to GTest version >=1.13.0 (#1501) @bdice
• Remove raft/matrix/matrix.cuh includes (#1498) @benfred
• Generate dataset of select_k times (#1497) @benfred
• Re-use memory pool between benchmark runs (#1495) @benfred
• Support CUDA 12.0 for pip wheels (#1489) @divyegala
• Update cupy dependency (#1488) @vyasr
• Enable sccache hits from local builds (#1478) @AyodeAwe
• Build wheels using new single image workflow (#1477) @vyasr
• Revert shared-action-workflows pin (#1475) @divyegala
• CAGRA: Separate graph index sorting functionality from prune function (#1471) @enp1s0
• Add generic reduction functions and separate reductions/warp_primitives (#1470) @akifcorduk
• [ENH] [FINAL] Header structure: combine all PRs into one (#1469) @ahendriksen
• use matrix::select_k in brute_force::knn call (#1463) @benfred
• Dropping Python 3.8 (#1454) @divyegala
• Fix linalg::map to work with non-power-of-2-sized types again (#1453) @ahendriksen
• [ENH] Enable building with clang (limit strict error checking to GCC) (#1452) @ahendriksen
• Remove usage of rapids-get-rapids-version-from-git (#1436) @jjacobelli
• Minor Updates to Sparse Structures (#1432) @divyegala
• Use nvtx3 includes. (#1431) @bdice
• Remove wheel pytest verbosity (#1424) @sevagh
• Add python bindings for matrix::select_k (#1422) @benfred
• Using raft::resources across raft::random (#1420) @cjnolet
• Generate build metrics report for test and benchmarks (#1414) @divyegala
• Update clang-format to 16.0.1. (#1412) @bdice
• Use ARC V2 self-hosted runners for GPU jobs (#1410) @jjacobelli
• Remove uses-setup-env-vars (#1406) @vyasr
• Resolve conflicts in auto-merger of branch-23.06 and branch-23.04 (#1403) @galipremsagar
• Adding base header-only conda package without cuda math libs (#1386) @cjnolet
• Fix IVF-PQ API to use device_vector_view (#1384) @lowener
• Branch 23.06 merge 23.04 (#1379) @vyasr
• Forward merge branch 23.04 into 23.06 (#1350) @cjnolet
• Fused L2 1-NN based on cutlass 3xTF32 / DMMA (#1118) @mdoijade