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.
We should figure out a good granularity for which to do this- whether it's by specialize a couple expensive functions in detail or just by specializing the user facing APIs for IVF-flat. The IVF-flat googletest, specifically, is at the top of the list of items taking a long time to compile (measuring ~20minutes on a single arch).
We should figure out a good granularity for which to do this- whether it's by specialize a couple expensive functions in
detail
or just by specializing the user facing APIs for IVF-flat. The IVF-flat googletest, specifically, is at the top of the list of items taking a long time to compile (measuring ~20minutes on a single arch).