felix-wang

Fix: build cuda extensions without nvidia runtime

1

The origin issue is from: https://github.com/pytorch/pytorch/issues/27222

chore: use stage host

2

DON'T MERGE THIS PR. It is raised for debugging only!

- [x] CI workflow - [ ] black and lint checking - [x] unittest - [ ] CD workflow - [ ] release pypi package - [ ] release executor

paper-reading-hm-ann

1

> HM-ANN enables billion-scale similarity search on a single machine without compression technologies. [HM-ANN: Efficient Billion-Point Nearest Neighbor Search on Heterogeneous Memory](https://papers.nips.cc/paper/2020/file/788d986905533aba051261497ecffcbb-Paper.pdf)

PQLite: bench SQL indexer

2

- sqlite - apsw (Another Python SQLite wrapper) https://github.com/rogerbinns/apsw - duckdb https://github.com/duckdb/duckdb - monetdb https://github.com/MonetDBSolutions/MonetDBe-Python references: - https://hal.inria.fr/hal-02556400/document

If the number fo valid data is small, we can directly apply bruteforce search over the subset data to speed up.