423A35C7/pytorch3d
1
0
Fork 0
mirror of https://github.com/facebookresearch/pytorch3d.git synced 2025-12-19 13:50:35 +08:00
Code Issues Packages Projects Releases Wiki Activity

remove requires_grad from random rotations

Browse Source 
Summary: Because rotations and (rotation) quaternions live on curved manifolds, it doesn't make sense to optimize them directly. Having a prominent option to require gradient on random ones may cause people to try, and isn't particularly useful.

Reviewed By: theschnitz

Differential Revision: D29160734

fbshipit-source-id: fc9e320672349fe334747c5b214655882a460a62
This commit is contained in:
Jeremy Reizenstein
2021-06-21 11:45:01 -07:00
committed by Facebook GitHub Bot
parent 31c448a95d
commit ce60d4b00e
2 changed files with 12 additions and 23 deletions
Download Patch File Download Diff File Expand all files Collapse all files

9
tests/test_rotation_conversions.py
View File

@@ -76,7 +76,8 @@ class TestRotationConversion(TestCaseMixin, unittest.TestCase):
def test_quat_grad_exists(self):
"""Quaternion calculations are differentiable."""
rotation = random_rotation(requires_grad=True)
rotation = random_rotation()
rotation.requires_grad = True
modified = quaternion_to_matrix(matrix_to_quaternion(rotation))
[g] = torch.autograd.grad(modified.sum(), rotation)
self.assertTrue(torch.isfinite(g).all())
@@ -131,7 +132,8 @@ class TestRotationConversion(TestCaseMixin, unittest.TestCase):
def test_euler_grad_exists(self):
"""Euler angle calculations are differentiable."""
rotation = random_rotation(dtype=torch.float64, requires_grad=True)
rotation = random_rotation(dtype=torch.float64)
rotation.requires_grad = True
for convention in self._all_euler_angle_conventions():
euler_angles = matrix_to_euler_angles(rotation, convention)
mdata = euler_angles_to_matrix(euler_angles, convention)
@@ -218,7 +220,8 @@ class TestRotationConversion(TestCaseMixin, unittest.TestCase):
def test_quaternion_application(self):
"""Applying a quaternion is the same as applying the matrix."""
quaternions = random_quaternions(3, torch.float64, requires_grad=True)
quaternions = random_quaternions(3, torch.float64)
quaternions.requires_grad = True
matrices = quaternion_to_matrix(quaternions)
points = torch.randn(3, 3, dtype=torch.float64, requires_grad=True)
transform1 = quaternion_apply(quaternions, points)