push sheeet

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

@@ -0,0 +1,65 @@
+{
+  lib,
+  buildPythonPackage,
+  fetchFromGitHub,
+  jaxtyping,
+  linear-operator,
+  mpmath,
+  scikit-learn,
+  scipy,
+  setuptools,
+  setuptools-scm,
+  torch,
+  pytestCheckHook,
+}:
+
+buildPythonPackage rec {
+  pname = "gpytorch";
+  version = "1.14";
+  pyproject = true;
+
+  src = fetchFromGitHub {
+    owner = "cornellius-gp";
+    repo = "gpytorch";
+    tag = "v${version}";
+    hash = "sha256-whZjqAs3nyjKMzAGi+OnyBtboq0UuV8m11A4IzkWtec=";
+  };
+
+  build-system = [
+    setuptools
+    setuptools-scm
+  ];
+
+  pythonRelaxDeps = [ "jaxtyping" ];
+
+  dependencies = [
+    jaxtyping
+    linear-operator
+    mpmath
+    scikit-learn
+    scipy
+    torch
+  ];
+
+  nativeCheckInputs = [ pytestCheckHook ];
+
+  pythonImportsCheck = [ "gpytorch" ];
+
+  disabledTests = [
+    # AssertionError on number of warnings emitted
+    "test_deprecated_methods"
+    # flaky numerical tests
+    "test_classification_error"
+    "test_matmul_matrix_broadcast"
+    "test_optimization_optimal_error"
+    # https://github.com/cornellius-gp/gpytorch/issues/2396
+    "test_t_matmul_matrix"
+  ];
+
+  meta = with lib; {
+    description = "Highly efficient and modular implementation of Gaussian Processes, with GPU acceleration";
+    homepage = "https://gpytorch.ai";
+    license = licenses.mit;
+    maintainers = with maintainers; [ veprbl ];
+  };
+}