Something went wrong. Try again.
What happens if I push all of nixpkgs
Something went wrong. Try again.
1.9 kB · 79 lines
Nix
at litex
1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980{ buildPythonPackage, embedding-reader, faiss, fetchFromGitHub, fire, fsspec, lib, numpy, pyarrow, pytestCheckHook, pythonRelaxDepsHook, pythonOlder}:
buildPythonPackage rec { pname = "autofaiss"; version = "2.15.5"; format = "setuptools";
disabled = pythonOlder "3.6";
src = fetchFromGitHub { owner = "criteo"; repo = pname; rev = "refs/tags/${version}"; hash = "sha256-IcAlvFlCERnw1UQWPRpSWpscOuPx0wd1MXOfoXZhvCU="; };
nativeBuildInputs = [ pythonRelaxDepsHook ];
pythonRemoveDeps = [ # The `dataclasses` packages is a python2-only backport, unnecessary in # python3. "dataclasses" # We call it faiss, not faiss-cpu. "faiss-cpu" ];
pythonRelaxDeps = [ # As of v2.15.4, autofaiss asks for fire<0.5 but we have fire v0.5.0 in # nixpkgs at the time of writing (2022-12-25). "fire" # As of v2.15.3, autofaiss asks for pyarrow<8 but we have pyarrow v9.0.0 in # nixpkgs at the time of writing (2022-12-15). "pyarrow" ];
propagatedBuildInputs = [ embedding-reader fsspec numpy faiss fire pyarrow ];
nativeCheckInputs = [ pytestCheckHook ];
disabledTests = [ # Attempts to spin up a Spark cluster and talk to it which doesn't work in # the Nix build environment. "test_build_partitioned_indexes" "test_index_correctness_in_distributed_mode_with_multiple_indices" "test_index_correctness_in_distributed_mode" "test_quantize_with_pyspark" ];
meta = with lib; { description = "Automatically create Faiss knn indices with the most optimal similarity search parameters"; homepage = "https://github.com/criteo/autofaiss"; changelog = "https://github.com/criteo/autofaiss/blob/${version}/CHANGELOG.md"; license = licenses.asl20; maintainers = with maintainers; [ samuela ]; };}