NVIDIA/raft v25.10.00

v25.10.00

Repository: NVIDIA/raft

Tag: v25.10.00

Published: 2025-10-08T17:49:50Z

Prerelease: no

Release notes:

🐛 Bug Fixes

• Workaround for an illegal memory access on SM 120 devices (#2821) @achirkin
• Fix sparse select_k: don't write beyond min(input_len, k) (#2814) @achirkin
• [BUG] Fix compilation error in matrix/detail/gather.cuh (#2811) @enp1s0
• Fix select_k for negative bfloat16 (#2799) @apivovarov
• Fix index types for coo kernels (#2793) @aamijar
• Fix the GEMM pointer mode setting (#2777) @achirkin
• Fix host_vector_policy issue (#2739) @viclafargue

📖 Documentation

• Fix UCX-Py mention to UCXX in docstring (#2804) @pentschev

🚀 New Features

• Update cutlass to a version that supports CUDA 13 (#2774) @robertmaynard

🛠️ Improvements

• Fix missed deps in update-version.sh (#2826) @AyodeAwe
• Empty commit to trigger a build (#2816) @msarahan
• Make warpsort kernels use the IEEE 754 bit representation for ordering (#2807) @achirkin
• Configure repo for automatic release notes generation (#2806) @AyodeAwe
• Support < 2 element arrays in rand_index/adjusted_rand_index (#2805) @jcrist
• update dependencies: use cuda-toolkit wheels (#2802) @jameslamb
• Use branch-25.10 again (#2800) @jameslamb
• Remove CMake find UCX package (#2798) @pentschev
• use dask-cuda[cu12, cu13] extras for wheel dependencies (#2797) @jameslamb
• Remove UCX-Py (#2791) @pentschev
• Update rapids-dependency-file-generator (#2790) @KyleFromNVIDIA
• Build and test with CUDA 13.0.0 (#2787) @jameslamb
• Fix template arg passing in adjusted_rand_index (#2785) @jinsolp
• Use build cluster in devcontainers (#2781) @trxcllnt
• Use rapids_cuda_enable_fatbin_compression (#2780) @robertmaynard
• Increase Dask tests verbosity in CI (#2779) @pentschev
• Update rapids_config to handle user defined branch name (#2778) @robertmaynard
• [REVIEW] Fix: skip default_allocation_limit() if unnecessary (#2775) @i-Pear
• Update rapids-build-backend to 0.4.1 (#2773) @KyleFromNVIDIA
• ci(labeler): update labeler action to @v5 (#2772) @gforsyth
• Register bfloat16/bfloat162 in util/vectorized.cuh (#2769) @apivovarov
• Use mdspan::index_type to Only Instantiate Specific Kernels (#2767) @tarang-jain
• Allow latest OS in devcontainers (#2759) @bdice
• Update build infra to support new branching strategy (#2751) @robertmaynard
• Use GCC 14 in conda builds. (#2708) @vyasr