rapidsai / raft

RAFT contains fundamental widely-used algorithms and primitives for machine learning and information retrieval. The algorithms are CUDA-accelerated and form building blocks for more easily writing high performance applications.

Home Page:https://docs.rapids.ai/api/raft/stable/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

[BUG] `raft-ann-bench` fails to benchmark the build time of `faiss_cpu_flat`.

fy opened this issue · comments

commented

Describe the bug

raft-ann-bench fails to benchmark the build time of faiss_cpu_flat. The index seems to build successfully, as the index file is saved, but the result file is not created.

Steps/Code to reproduce bug

docker run --rm -u $(id -u) --gpus all --entrypoint /bin/bash --workdir /data/benchmarks -v .:/data/benchmarks rapidsai/raft-ann-bench:24.02a-cuda11.8-py3.10 -c 'eval "$(conda shell.bash hook)"
python -m raft-ann-bench.run --build --dataset glove-100-inner --algorithms faiss_cpu_flat --force'

Expected behavior

  • An index file is built and saved as ./datasets/glove-100-inner/index/faiss_cpu_flat
  • Missing: A result file is produced ./datasets/glove-100-inner/result/build/faiss_cpu_flat,base.json

Additional context
Error message

Error occurred running benchmark: Command '['/opt/conda/bin/ann/FAISS_CPU_FLAT_ANN_BENCH', '--build', '--data_prefix=/data/benchmarks/datasets/', '--benchmark_out_format=json', '--benchmark_counters_tabular=true', '--benchmark_out=/data/benchmarks/datasets/glove-100-inner/result/build/faiss_cpu_flat,base.json.lock', '--raft_log_level=3', '--force', 'glove-100-inner_faiss_cpu_flat,base,k10,bs10000_a3f4f69a-c183-11ee-ad15-0242ac110003.json']' died with <Signals.SIGSEGV: 11>.
commented

Could the issue be due to vector dimension incompatibility? I reran the benchmark with the faiss_cpu_flat algorithm for the wiki_all_1M dataset, and I did not encounter any errors.