depot/third_party/nixpkgs/pkgs/development/python-modules/pymanopt/default.nix

53 lines
1.2 KiB
Nix
Raw Normal View History

{ lib
, fetchFromGitHub
, buildPythonPackage
, numpy
, scipy
, torch
, autograd
, nose2
, matplotlib
, tensorflow
}:
buildPythonPackage rec {
pname = "pymanopt";
version = "2.0.1";
src = fetchFromGitHub {
owner = pname;
repo = pname;
rev = "refs/tags/${version}";
sha256 = "sha256-VwCUqKI1PkR8nUVaa73bkTw67URKPaza3VU9g+rB+Mg=";
};
propagatedBuildInputs = [ numpy scipy torch ];
nativeCheckInputs = [ nose2 autograd matplotlib tensorflow ];
checkPhase = ''
runHook preCheck
# upstream themselves seem unsure about the robustness of these
# tests - see https://github.com/pymanopt/pymanopt/issues/219
grep -lr 'test_second_order_function_approximation' tests/ | while read -r fn ; do
substituteInPlace "$fn" \
--replace \
'test_second_order_function_approximation' \
'dont_test_second_order_function_approximation'
done
nose2 tests -v
runHook postCheck
'';
pythonImportsCheck = [ "pymanopt" ];
meta = {
description = "Python toolbox for optimization on Riemannian manifolds with support for automatic differentiation";
homepage = "https://www.pymanopt.org/";
license = lib.licenses.bsd3;
maintainers = with lib.maintainers; [ yl3dy ];
};
}