Something went wrong. Try again.
What happens if I push all of nixpkgs
Something went wrong. Try again.
1.8 kB · 91 lines
Nix
1234567891011121314151617181920212223242526272829303132333435363738394041424344454647484950515253545556575859606162636465666768697071727374757677787980818283848586878889909192{ lib, buildPythonPackage, fetchFromGitHub,
# build-system setuptools,
# dependencies dm-tree, etils, numpy, tabulate, wrapt,
# tests click, docutils, keras, pytestCheckHook, tensorflow, tensorflow-datasets, tf-keras,}:
buildPythonPackage rec { pname = "dm-sonnet"; version = "2.0.2"; pyproject = true;
src = fetchFromGitHub { owner = "deepmind"; repo = "sonnet"; tag = "v${version}"; hash = "sha256-WkloUbqSyPG3cbLG8ktsjdluACkCbUZ7t6rYWst8rs8="; };
build-system = [ setuptools ];
dependencies = [ dm-tree etils numpy tabulate wrapt ] ++ etils.optional-dependencies.epath;
optional-dependencies = { tensorflow = [ tensorflow ]; };
nativeCheckInputs = [ click docutils keras pytestCheckHook tensorflow tensorflow-datasets tf-keras ];
# ImportError: `keras.optimizers.legacy` is not supported in Keras 3 preCheck = '' export TF_USE_LEGACY_KERAS=True '';
disabledTests = [ # AssertionError: 2 != 0 : 2 doctests failed "test_doctest_sonnet.functional"
# AssertionError: Not equal to tolerance "testComputationAgainstNumPy1"
# tensorflow.python.framework.errors_impl.InvalidArgumentError: cannot compute MatMul as input #1(zero-based) was expected to be a float tensor but is a half tensor [Op:MatMul] "testComputationAgainstNumPy0" "testComputationAgainstNumPy1" ];
pythonImportsCheck = [ "sonnet" ];
meta = { description = "Library for building neural networks in TensorFlow"; homepage = "https://github.com/deepmind/sonnet"; changelog = "https://github.com/google-deepmind/sonnet/releases/tag/v${version}"; license = lib.licenses.asl20; maintainers = with lib.maintainers; [ onny ]; };}