NVIDIA/cuvs v26.06.00

v26.06.00

Repository: NVIDIA/cuvs

Tag: v26.06.00

Published: 2026-06-04T00:21:20Z

Prerelease: no

Release notes:

What's Changed

🚨 Breaking Changes

• Default to static linking of libcudart by @bdice in https://github.com/rapidsai/cuvs/pull/1627
• Remove JIT+LTO fragment database by @KyleFromNVIDIA in https://github.com/rapidsai/cuvs/pull/1927
• Use static cudart by @KyleFromNVIDIA in https://github.com/rapidsai/cuvs/pull/1931
• Always build with JIT+LTO by @KyleFromNVIDIA in https://github.com/rapidsai/cuvs/pull/1923
• Migrate RMM usage to CCCL MR design by @bdice in https://github.com/rapidsai/cuvs/pull/1990
• Exposition of KMeans param object for PQ in C++ by @lowener in https://github.com/rapidsai/cuvs/pull/2005
• [Cleanup] Combine Batched and Regular KMeans Impl by @tarang-jain in https://github.com/rapidsai/cuvs/pull/2015
• Preserve input memory location for NN Descent by @jinsolp in https://github.com/rapidsai/cuvs/pull/1928

🐛 Bug Fixes

• Fix CCCL compilation error by @viclafargue in https://github.com/rapidsai/cuvs/pull/1963
• Forward-merge release/26.04 into main by @KyleFromNVIDIA in https://github.com/rapidsai/cuvs/pull/1971
• Forward-merge release/26.04 into main by @KyleFromNVIDIA in https://github.com/rapidsai/cuvs/pull/1980
• Remove dangling pointers in JIT Fragments by @divyegala in https://github.com/rapidsai/cuvs/pull/1988
• Add head_rev to cuvs recipe by @KyleFromNVIDIA in https://github.com/rapidsai/cuvs/pull/1993
• Fix potential OOB access in CAGRA search when graph size < dataset size by @irina-resh-nvda in https://github.com/rapidsai/cuvs/pull/1780
• Fix MG kmeans intertia_check n_iters by @aamijar in https://github.com/rapidsai/cuvs/pull/2020
• Fix cuvs_bench pytest pareto assert by @aamijar in https://github.com/rapidsai/cuvs/pull/2027
• Fix nightly build matrix by @KyleFromNVIDIA in https://github.com/rapidsai/cuvs/pull/2054
• Fix vulnerable index deserialization by @lowener in https://github.com/rapidsai/cuvs/pull/2068
• Fix symbol export kmeans by @aamijar in https://github.com/rapidsai/cuvs/pull/2070
• Fix argmin/argmax based on the distance type by @achirkin in https://github.com/rapidsai/cuvs/pull/2016
• Remove unneeded request for CUDA device link phase by @robertmaynard in https://github.com/rapidsai/cuvs/pull/2077
• Update Faiss and DiskANN Patch to Use C++20 by @tarang-jain in https://github.com/rapidsai/cuvs/pull/1796
• Fix brute force Rust index dataset lifetime by @yan-zaretskiy in https://github.com/rapidsai/cuvs/pull/2083
• Fix segfault cuvs bench by @aamijar in https://github.com/rapidsai/cuvs/pull/2088
• Fix cagra::optimize modifying the state of raft::resources by @achirkin in https://github.com/rapidsai/cuvs/pull/2103
• Add direct target dependency when embedding fatbins by @KyleFromNVIDIA in https://github.com/rapidsai/cuvs/pull/2106
• Fix check for PQ vectorized load by @lowener in https://github.com/rapidsai/cuvs/pull/2107
• Fix workspace usage by @mfoerste4 in https://github.com/rapidsai/cuvs/pull/2135
• Add missing visibility controls in IVF SQ by @divyegala in https://github.com/rapidsai/cuvs/pull/2141

📖 Documentation

• Elaborate on fragment architecture in JIT+LTO documentation by @KyleFromNVIDIA in https://github.com/rapidsai/cuvs/pull/1991
• Add Cluster and Distance sections to C documentation by @lowener in https://github.com/rapidsai/cuvs/pull/1955
• Adding CAGRA merge to the documentation by @viclafargue in https://github.com/rapidsai/cuvs/pull/1942
• [Doc Update] CAGRA Memory Footprint by @singhmanas1 in https://github.com/rapidsai/cuvs/pull/1300
• Align docs with pluggable benchmark API by @jnke2016 in https://github.com/rapidsai/cuvs/pull/1891
• Add docs for cagra mem usage with NN Descent build algo by @jinsolp in https://github.com/rapidsai/cuvs/pull/2000
• Fix minor typos in `cuvs-bench` source build docs by @jrbourbeau in https://github.com/rapidsai/cuvs/pull/2006
• Fix cuvs-bench docker images in docs by @jrbourbeau in https://github.com/rapidsai/cuvs/pull/2003
• Update JIT+LTO guide to reflect new automatic embedding system by @KyleFromNVIDIA in https://github.com/rapidsai/cuvs/pull/2045
• [Docs] Convert Sphinx docs to Fern by @cjnolet in https://github.com/rapidsai/cuvs/pull/2067
• Add UDF Usage and Developer docs by @divyegala in https://github.com/rapidsai/cuvs/pull/2030
• [DOC] Adding API guides for core cuVS types by @cjnolet in https://github.com/rapidsai/cuvs/pull/2117

🚀 New Features

• [REVIEW] Add L1 support to NN-Descent by @yan-zaretskiy in https://github.com/rapidsai/cuvs/pull/1898
• PCA C and Python API by @aamijar in https://github.com/rapidsai/cuvs/pull/1987
• Introduce UDF Architecture by @divyegala in https://github.com/rapidsai/cuvs/pull/1804
• JIT LTO Cagra Search by @divyegala in https://github.com/rapidsai/cuvs/pull/1807
• Expose supported brute force metrics in all_neighbors by @jinsolp in https://github.com/rapidsai/cuvs/pull/1827
• [REVIEW] Generalize and improve cagra::optimize by @mfoerste4 in https://github.com/rapidsai/cuvs/pull/1830
• IVF-SQ C++ API by @viclafargue in https://github.com/rapidsai/cuvs/pull/1865

🛠️ Improvements

• Use PQ API in CAGRA-Q + SCANN by @lowener in https://github.com/rapidsai/cuvs/pull/1746
• Speed up recall calculation in cuVS Bench for large top-K by @jamxia155 in https://github.com/rapidsai/cuvs/pull/1816
• Update codespell Version in pre-commit-config by @tarang-jain in https://github.com/rapidsai/cuvs/pull/1920
• Forward-merge release/26.04 into main by @gforsyth in https://github.com/rapidsai/cuvs/pull/1936
• Refactor StaticFatbinFragmentEntry to use tags by @KyleFromNVIDIA in https://github.com/rapidsai/cuvs/pull/1970
• Replace cudaMemcpy2DAsync Calls with raft::copy_matrix by @tarang-jain in https://github.com/rapidsai/cuvs/pull/1976
• update pip devcontainers' base image tags by @trxcllnt in https://github.com/rapidsai/cuvs/pull/1985
• Refactor instantiation matrices to generate at build time by @KyleFromNVIDIA in https://github.com/rapidsai/cuvs/pull/1984
• Add option to enable "sve" optimization level on armv9 by @LizYou in https://github.com/rapidsai/cuvs/pull/1121
• Improve cuvs-bench doc and add executable dir option by @tfeher in https://github.com/rapidsai/cuvs/pull/681
• Enforce type safety in JIT+LTO launcher by @KyleFromNVIDIA in https://github.com/rapidsai/cuvs/pull/1997
• Add KDE kernel by @Intron7 in https://github.com/rapidsai/cuvs/pull/1915
• Coderabbit integration by @benfred in...