python3Packages.tensorly: init at 0.4.5

pkgs/development/python-modules/tensorly/default.nix

@@ -0,0 +1,45 @@
+{ lib
+, buildPythonPackage
+, fetchFromGitHub
+, pytest
+, nose
+, isPy27
+, numpy
+, scipy
+, sparse
+, pytorch
+}:
+
+buildPythonPackage rec {
+  pname = "tensorly";
+  version = "0.4.5";
+  disabled = isPy27;
+
+  src = fetchFromGitHub {
+    owner = pname;
+    repo = pname;
+    rev = version;
+    sha256 = "1ml91yaxwx4msisxbm92yf22qfrscvk58f3z2r1jhi96pw2k4i7x";
+  };
+
+  propagatedBuildInputs = [ numpy scipy sparse ];
+  checkInputs = [ pytest nose pytorch ];
+  # also has a cupy backend, but the tests are currently broken
+  # (e.g. attempts to access cupy.qr instead of cupy.linalg.qr)
+  # and this backend also adds a non-optional CUDA dependence,
+  # as well as tensorflow and mxnet backends, but the tests don't
+  # seem to exercise these backend by default
+
+  checkPhase = ''
+    runHook preCheck
+    nosetests -e "test_cupy"
+    runHook postCheck
+  '';
+
+  meta = with lib; {
+    description = "Tensor learning in Python";
+    homepage = https://tensorly.org/;
+    license = licenses.bsd3;
+    maintainers = [ maintainers.bcdarwin ];
+  };
+}

pkgs/top-level/python-packages.nix

@@ -6581,6 +6581,8 @@ in {
     cudaSupport = true;
   };
 
+  tensorly = callPackage ../development/python-modules/tensorly { };
+
   tflearn = callPackage ../development/python-modules/tflearn { };
 
   simpleai = callPackage ../development/python-modules/simpleai { };