Lucene: cuvsCagraSearch returned 0 (Reason=cudaErrorInvalidValue:invalid argument)
Open
@imotov is already working on this.
Since Aug 17, 2026.
bug
Lucene
non-breaking
- Dominant language
- Cuda
- Stars
- 854
- Forks
- 236
- Avg merge
- 3d 3h
- Merged PRs (30d)
- 62
Description
Issue:
[ERROR] com.nvidia.cuvs.lucene.TestCuVSRandomizedVectorSearch.testVectorSearch -- Time elapsed: 0.157 s <<< ERROR!
java.lang.RuntimeException:
cuvsCagraSearch returned 0[CUDA error encountered at: file=/home/narang/code/cuvs/cpp/src/neighbors/detail/cagra/search_single_cta_kernel-inl.cuh line=2207: call='cudaPeekAtLastError()', Reason=cudaErrorInvalidValue:invalid argument
Obtained 8 stack frames
NVIDIA/cuvs-lucene#1 in /home/narang/code/cuvs/cpp/build/c/libcuvs_c.so: raft::cuda_error::cuda_error(std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&) +0x6e [0x77096bf4f90e]
NVIDIA/cuvs-lucene#2 in /home/narang/code/cuvs/cpp/build/libcuvs.so(+0x1897edf) [0x77088f297edf]
NVIDIA/cuvs-lucene#3 in /home/narang/code/cuvs/cpp/build/libcuvs.so: cuvs::neighbors::cagra::detail::single_cta_search::search<float, unsigned int, float, cuvs::neighbors::filtering::none_sample_filter, unsigned int, unsigned int>::operator()(raft::resources const&, cuda::std::__4::mdspan<unsigned int const, cuda::std::__4::extents<long, 18446744073709551615ul, 18446744073709551615ul>, cuda::std::__4::layout_right, raft::host_device_accessor<cuda::std::__4::default_accessor<unsigned int const>, (raft::memory_type)2> >, std::optional<cuda::std::__4::mdspan<unsigned int const, cuda::std::__4::extents<long, 18446744073709551615ul>, cuda::std::__4::layout_right, raft::host_device_accessor<cuda::std::__4::default_accessor<unsigned int const>, (raft::memory_type)2> > >, unsigned int*, float*, float const*, unsigned int, unsigned int const*, unsigned int*, unsigned int, cuvs::neighbors::filtering::none_sample_filter) +0x16f [0x77088f2df0bf]
NVIDIA/cuvs-lucene#4 in /home/narang/code/cuvs/cpp/build/libcuvs.so(+0x18cc812) [0x77088f2cc812]
NVIDIA/cuvs-lucene#5 in /home/narang/code/cuvs/cpp/build/libcuvs.so: void cuvs::neighbors::cagra::search<float, unsigned int, unsigned int>(raft::resources const&, cuvs::neighbors::cagra::search_params const&, cuvs::neighbors::cagra::index<float, unsigned int> const&, cuda::std::__4::mdspan<float const, cuda::std::__4::extents<long, 18446744073709551615ul, 18446744073709551615ul>, cuda::std::__4::layout_right, raft::host_device_accessor<cuda::std::__4::default_accessor<float const>, (raft::memory_type)2> >, cuda::std::__4::mdspan<unsigned int, cuda::std::__4::extents<long, 18446744073709551615ul, 18446744073709551615ul>, cuda::std::__4::layout_right, raft::host_device_accessor<cuda::std::__4::default_accessor<unsigned int>, (raft::memory_type)2> >, cuda::std::__4::mdspan<float, cuda::std::__4::extents<long, 18446744073709551615ul, 18446744073709551615ul>, cuda::std::__4::layout_right, raft::host_device_accessor<cuda::std::__4::default_accessor<float>, (raft::memory_type)2> >, cuvs::neighbors::filtering::base_filter const&) +0x3e5 [0x77088f2dbe15]
NVIDIA/cuvs-lucene#6 in /home/narang/code/cuvs/cpp/build/c/libcuvs_c.so(+0xab40b) [0x77096bf8040b]
NVIDIA/cuvs-lucene#7 in /home/narang/code/cuvs/cpp/build/c/libcuvs_c.so: cuvsCagraSearch +0x51 [0x77096bf817b1]
NVIDIA/cuvs-lucene#8 in [0x7709983cf2e2]
]
at __randomizedtesting.SeedInfo.seed([55C8B265D329C693:C1C1A0838FF56B00]:0)
at com.nvidia.cuvs.internal.common.Util.checkCuVSError(Util.java:87)
at com.nvidia.cuvs.internal.CagraIndexImpl.search(CagraIndexImpl.java:309)
at com.nvidia.cuvs.lucene.CuVS2510GPUVectorsReader.search(CuVS2510GPUVectorsReader.java:453)
.
.
.
To reproduce this issue do:
mvn clean test -Dtest=TestCuVSRandomizedVectorSearch#testVectorSearch -Dtests.seed=55C8B265D329C693
Contributor guide
First steps
- Read the whole issue, then the project's contributing guide.
- Comment on the issue to say you are picking it up — it saves two people doing the same work.
- Fork the repository and make your change on a branch.
- Open a pull request that references the issue number.
Assessment
This issue has not been assessed yet.