Corey J. Nolet

Documentation should include comprehensive examples of how to use the different constructs

8

I'm looking through the code to figure out how to use the DenseVector and the UniformTimeIndex, for instance, and I've had parse quite a bit of code to figure them...

Investigate integration of Scalation primitives and expose them to RDDs

2

Scalation, or, Scalable Simulation, has a pretty robust set of primitives for modeling time series data. They are porting a few of the more complicated algorithms from R to Scala....

Will the default CNN models generalize well to non-english languages like persian or chinese?

2

I ask this because many of the other proposal methods (like edge boxes) the classifiers generalize well across different category types.

Would be nice if a LICENSE.txt file was created for users.

[FEA] Increase CAGRA's max top-k

4

Summarizing an offline discussion w/ @anaruse (Akira, please feel free to correct any mistakes): In CAGRA's single-CTA algorithm, the internal top-k size is capped at 512 (and `k` must be...

After we build our C API (for CAGRA), we should build a Rust API wrapper around it as a proof of concept.

This is a very useful calculation for many downstream database systems to be able to schedule index builds against the GPU.

[FEA] Primitives for pre-filtering and post-filtering

2

The IVF methods currently accept pre-filtering functions that can be applied during scan but still need optimized primitives that can allow users to efficiently express and perform the filtering logic....

This workaround was accidentally reverted during [this commit](https://github.com/rapidsai/raft/commit/dc755907675e36e026535cf3bb48dc19d2e42470) but the additional memory copy is needed for proper results in CUDA 12.2-12.4.

This is a tracker/placeholder issue to support this feature as a follow-on to the initial RAFT IVFPQ support in FAISS.