Skip to content

Add type annotations, parameterize a test in gnm_common_test.py. #54

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
copybara-service merged 1 commit into from
Jul 27, 2026
+52 −36
Merged

Add type annotations, parameterize a test in gnm_common_test.py. #54

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
Binary file modified gnm/shape/data/versions/v3_0/gnm_head.npz
View file
Open in desktop
Binary file not shown.
87 changes: 51 additions & 36 deletions gnm/shape/gnm_common_test.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,13 +12,16 @@
# See the License for the specific language governing permissions and
# limitations under the License.

"""Tests for the backend-agnostic GNM math."""

from collections.abc import Sequence

from absl.testing import absltest
from absl.testing import parameterized
from etils import enp
from gnm.shape import gnm_common
import numpy as np


# The array backends supported by the GNM math, as (name, xnp module) pairs.
# Every test is run against all of them so that backend-specific regressions
# are caught. The xnp modules mirror those used by the per-backend GNM classes
Expand All @@ -35,21 +38,21 @@ class ArrayHelpersTest(parameterized.TestCase):
"""Tests for the backend-agnostic array helpers."""

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_take_selects_along_axis(self, xnp):
def test_take_selects_along_axis(self, xnp: enp.NpModule):
array = xnp.asarray([[0, 1, 2], [3, 4, 5], [6, 7, 8]], dtype=xnp.float32)
indices = xnp.asarray([0, 2], dtype=xnp.int32)
taken = gnm_common.take(array, indices, axis=0, xnp=xnp)
np.testing.assert_array_equal(np.asarray(taken), [[0, 1, 2], [6, 7, 8]])

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_take_selects_along_last_axis(self, xnp):
def test_take_selects_along_last_axis(self, xnp: enp.NpModule):
array = xnp.asarray([[0, 1, 2], [3, 4, 5]], dtype=xnp.float32)
indices = xnp.asarray([2, 0], dtype=xnp.int32)
taken = gnm_common.take(array, indices, axis=-1, xnp=xnp)
np.testing.assert_array_equal(np.asarray(taken), [[2, 0], [5, 3]])

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_eye_returns_identity(self, xnp):
def test_eye_returns_identity(self, xnp: enp.NpModule):
reference = xnp.asarray([0.0, 0.0, 0.0, 0.0], dtype=xnp.float32)
identity = gnm_common.eye(
3, dtype=xnp.float32, reference_array=reference, xnp=xnp
Expand All @@ -58,7 +61,7 @@ def test_eye_returns_identity(self, xnp):
np.testing.assert_array_equal(np.asarray(identity), np.eye(3))

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_reshape_with_batch_dims(self, xnp):
def test_reshape_with_batch_dims(self, xnp: enp.NpModule):
reference = xnp.asarray(np.zeros((2, 5, 3)), dtype=xnp.float32)
array = xnp.asarray(np.arange(2 * 6), dtype=xnp.float32)
reshaped = gnm_common.reshape_with_batch_dims(
Expand All @@ -70,7 +73,7 @@ def test_reshape_with_batch_dims(self, xnp):
self.assertEqual(tuple(reshaped.shape), (2, 6))

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_zeros_with_batch_dims(self, xnp):
def test_zeros_with_batch_dims(self, xnp: enp.NpModule):
reference = xnp.asarray(np.ones((4, 5, 3)), dtype=xnp.float32)
zeros = gnm_common.zeros_with_batch_dims(
reference,
Expand All @@ -86,36 +89,42 @@ class AxisAngleToRotationMatrixTest(parameterized.TestCase):
"""Tests for axis_angle_to_rotation_matrix."""

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_zero_returns_identity(self, xnp):
def test_zero_returns_identity(self, xnp: enp.NpModule):
axis_angle = xnp.asarray([0.0, 0.0, 0.0], dtype=xnp.float32)
matrix = np.asarray(gnm_common.axis_angle_to_rotation_matrix(axis_angle))
np.testing.assert_allclose(matrix, np.eye(3), atol=1e-5)

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_axis_is_invariant(self, xnp):
@parameterized.product(
xnp=tuple(xnp for _, xnp in _XNP_BACKENDS),
case=(
([np.pi / 2, 0.0, 0.0], [1.0, 0.0, 0.0]),
([0.0, np.pi / 2, 0.0], [0.0, 1.0, 0.0]),
([0.0, 0.0, np.pi / 2], [0.0, 0.0, 1.0]),
),
)
def test_axis_is_invariant(
self,
xnp: enp.NpModule,
case: tuple[Sequence[float], Sequence[float]],
) -> None:
# Rotating about an axis leaves that axis unchanged.
cases = (
([np.pi / 2, 0.0, 0.0], [1.0, 0.0, 0.0]),
([0.0, np.pi / 2, 0.0], [0.0, 1.0, 0.0]),
([0.0, 0.0, np.pi / 2], [0.0, 0.0, 1.0]),
axis_angle, axis = case
matrix = np.asarray(
gnm_common.axis_angle_to_rotation_matrix(
xnp.asarray(axis_angle, dtype=xnp.float32)
)
)
for axis_angle, axis in cases:
matrix = np.asarray(
gnm_common.axis_angle_to_rotation_matrix(
xnp.asarray(axis_angle, dtype=xnp.float32)
)
)
np.testing.assert_allclose(matrix @ np.asarray(axis), axis, atol=1e-5)
np.testing.assert_allclose(matrix @ np.asarray(axis), axis, atol=1e-5)

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_ninety_degrees_about_z(self, xnp):
def test_ninety_degrees_about_z(self, xnp: enp.NpModule):
axis_angle = xnp.asarray([0.0, 0.0, np.pi / 2], dtype=xnp.float32)
matrix = np.asarray(gnm_common.axis_angle_to_rotation_matrix(axis_angle))
expected = np.array([[0.0, -1.0, 0.0], [1.0, 0.0, 0.0], [0.0, 0.0, 1.0]])
np.testing.assert_allclose(matrix, expected, atol=1e-5)

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_matrix_is_orthonormal_with_unit_determinant(self, xnp):
def test_matrix_is_orthonormal_with_unit_determinant(self, xnp: enp.NpModule):
rng = np.random.default_rng(0)
axis_angle = rng.uniform(-3.0, 3.0, size=(8, 3))
matrices = np.asarray(
Expand All @@ -131,7 +140,7 @@ def test_matrix_is_orthonormal_with_unit_determinant(self, xnp):
np.testing.assert_allclose(np.linalg.det(matrices), np.ones(8), atol=1e-4)

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_preserves_batch_shape(self, xnp):
def test_preserves_batch_shape(self, xnp: enp.NpModule):
axis_angle = xnp.asarray(np.zeros((2, 4, 3)), dtype=xnp.float32)
matrices = gnm_common.axis_angle_to_rotation_matrix(axis_angle)
self.assertEqual(tuple(matrices.shape), (2, 4, 3, 3))
Expand All @@ -147,7 +156,7 @@ def setUp(self):
self.parents = [0, 0]

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_rest_pose_places_joints_at_bind_positions(self, xnp):
def test_rest_pose_places_joints_at_bind_positions(self, xnp: enp.NpModule):
transforms = np.asarray(
gnm_common.joint_transforms_world(
xnp.asarray(self.joints, dtype=xnp.float32),
Expand All @@ -166,7 +175,7 @@ def test_rest_pose_places_joints_at_bind_positions(self, xnp):
)

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_global_translation_shifts_all_joints(self, xnp):
def test_global_translation_shifts_all_joints(self, xnp: enp.NpModule):
transforms = np.asarray(
gnm_common.joint_transforms_world(
xnp.asarray(self.joints, dtype=xnp.float32),
Expand All @@ -182,7 +191,7 @@ def test_global_translation_shifts_all_joints(self, xnp):
)

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_root_rotation_propagates_to_child(self, xnp):
def test_root_rotation_propagates_to_child(self, xnp: enp.NpModule):
# Rotate the root 90 degrees about z; the child at (1, 0, 0) maps to
# (0, 1, 0).
rotations = xnp.asarray(
Expand Down Expand Up @@ -213,7 +222,7 @@ def setUp(self):
self.vertices = [[0.5, 0.0, 0.0], [0.0, 0.5, 0.0], [1.0, 1.0, 0.0]]

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_rest_pose_is_identity(self, xnp):
def test_rest_pose_is_identity(self, xnp: enp.NpModule):
posed = np.asarray(
gnm_common.linear_blend_skinning(
xnp.asarray(self.vertices, dtype=xnp.float32),
Expand All @@ -228,7 +237,7 @@ def test_rest_pose_is_identity(self, xnp):
np.testing.assert_allclose(posed, self.vertices, atol=1e-5)

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_global_translation_translates_vertices(self, xnp):
def test_global_translation_translates_vertices(self, xnp: enp.NpModule):
translation = [10.0, 0.0, 0.0]
posed = np.asarray(
gnm_common.linear_blend_skinning(
Expand All @@ -245,7 +254,7 @@ def test_global_translation_translates_vertices(self, xnp):
)

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_preserves_batch_shape(self, xnp):
def test_preserves_batch_shape(self, xnp: enp.NpModule):
batch_vertices = np.array(np.broadcast_to(self.vertices, (4, 3, 3)))
batch_joints = np.array(np.broadcast_to(self.joints, (4, 2, 3)))
posed = gnm_common.linear_blend_skinning(
Expand All @@ -263,7 +272,9 @@ class BindPoseTest(parameterized.TestCase):
"""Tests for the bind-pose vertex and joint helpers."""

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_vertex_positions_none_params_return_template(self, xnp):
def test_vertex_positions_none_params_return_template(
self, xnp: enp.NpModule
):
template = xnp.asarray(np.ones((3, 3)), dtype=xnp.float32)
identity_basis = xnp.asarray(np.zeros((2, 3, 3)), dtype=xnp.float32)
expression_basis = xnp.asarray(np.zeros((2, 3, 3)), dtype=xnp.float32)
Expand All @@ -273,7 +284,9 @@ def test_vertex_positions_none_params_return_template(self, xnp):
np.testing.assert_allclose(np.asarray(result), np.ones((3, 3)), atol=1e-5)

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_vertex_positions_apply_identity_and_expression(self, xnp):
def test_vertex_positions_apply_identity_and_expression(
self, xnp: enp.NpModule
):
template = xnp.asarray(np.zeros((2, 3)), dtype=xnp.float32)
# (I=1, V=2, 3)
identity_basis = xnp.asarray(
Expand All @@ -294,7 +307,9 @@ def test_vertex_positions_apply_identity_and_expression(self, xnp):
np.testing.assert_allclose(np.asarray(result), expected, atol=1e-5)

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_joint_positions_none_identity_returns_template(self, xnp):
def test_joint_positions_none_identity_returns_template(
self, xnp: enp.NpModule
):
template = xnp.asarray(
[[0.0, 0.0, 0.0], [1.0, 0.0, 0.0]], dtype=xnp.float32
)
Expand All @@ -305,7 +320,7 @@ def test_joint_positions_none_identity_returns_template(self, xnp):
)

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_joint_positions_apply_identity(self, xnp):
def test_joint_positions_apply_identity(self, xnp: enp.NpModule):
template = xnp.asarray(np.zeros((2, 3)), dtype=xnp.float32)
# (I=1, J=2, 3)
joint_basis = xnp.asarray(
Expand All @@ -323,7 +338,7 @@ class PoseCorrectivesTest(parameterized.TestCase):
"""Tests for compute_pose_correctives."""

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_none_inputs_return_zeros(self, xnp):
def test_none_inputs_return_zeros(self, xnp: enp.NpModule):
template = xnp.asarray(np.ones((3, 3)), dtype=xnp.float32)
result = gnm_common.compute_pose_correctives(
None, None, template, num_joints=2, num_vertices=3
Expand All @@ -332,7 +347,7 @@ def test_none_inputs_return_zeros(self, xnp):
np.testing.assert_array_equal(np.asarray(result), np.zeros((3, 3)))

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_zero_rotations_produce_no_correctives(self, xnp):
def test_zero_rotations_produce_no_correctives(self, xnp: enp.NpModule):
template = xnp.asarray(np.ones((3, 3)), dtype=xnp.float32)
regressor = xnp.asarray(np.ones((2 * 9, 3 * 3)), dtype=xnp.float32)
result = gnm_common.compute_pose_correctives(
Expand All @@ -346,7 +361,7 @@ def test_zero_rotations_produce_no_correctives(self, xnp):
np.testing.assert_allclose(np.asarray(result), np.zeros((3, 3)), atol=1e-5)

@parameterized.named_parameters(*_XNP_BACKENDS)
def test_preserves_batch_shape(self, xnp):
def test_preserves_batch_shape(self, xnp: enp.NpModule):
template = xnp.asarray(np.ones((3, 3)), dtype=xnp.float32)
regressor = xnp.asarray(np.ones((2 * 9, 3 * 3)), dtype=xnp.float32)
result = gnm_common.compute_pose_correctives(
Expand Down
1 change: 1 addition & 0 deletions gnm/shape/gnm_landmarks_test.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -15,6 +15,7 @@
"""Tests for validating and loading GNM landmarks configurations."""

from absl.testing import absltest

from gnm.shape import gnm_landmarks
from gnm.shape.data.versions import gnm_specs
import numpy as np
Expand Down
Loading