diskann-quantization 0.48.0

DiskANN is a fast approximate nearest neighbor search library for high dimensional data
Documentation

Quantization

Refer to the crate level documentation for usage of the quantization crate.

This README refers to building instructions when using the flatbuffers-build feature of this crate, which compiles the files in the schemas directory and copies the contents into src/flatbuffers.

Building with flatbuffers-build

The flatbuffers compiler flatc needs to be installed and the environment variable FLATC_EXE must be defined with the path to the executable.

The executable can be downloaded from the official release pages: https://github.com/google/flatbuffers/releases/tag/v25.2.10

SHA-512 sums of the zip directories for v25.2.10 are as follows.

  • 6a20c2fc4e4e094574a0fd064f79a374eb9e6abba9e49d4543ec384b056725f6ca9f7823ba5952fcfa40e31a56a4e25baa659415d94edd69a7a978942577c579 Linux.flatc.binary.clang++-18.zip
  • 8784aae9f7984fdf5685e3944787bc547ca3a8bccefa4ba33efbe73960ebb6c94c2061d251dcc00e683133e65c8f47833195e0293415bc8abbd7b5aab4419714 Windows.flatc.binary.zip