add testing.assert_close

* fix bugs in test_spline.py Signed-off-by: Xiangyu Chen <[email protected]> * add testing Signed-off-by: Xiangyu Chen <[email protected]> * add testing.rst Signed-off-by: Xiangyu Chen <[email protected]> * add testing Signed-off-by: Xiangyu Chen <[email protected]> * Update __init__.py Signed-off-by: Xiangyu Chen <[email protected]> * Update test_spline.py Signed-off-by: Xiangyu Chen <[email protected]> * Update test_spline.py Signed-off-by: Xiangyu Chen <[email protected]> * rename the assert_clase * Update test_spline.py --------- Signed-off-by: Xiangyu Chen <[email protected]> Co-authored-by: Chen Wang <[email protected]> Co-authored-by: Chen Wang <[email protected]>
HongLouyemeng · Aug 19, 2023 · c71c6fb · c71c6fb
1 parent 6421ae2
commit c71c6fb
Show file tree

Hide file tree

Showing 7 changed files with 89 additions and 36 deletions.
diff --git a/docs/source/index.rst b/docs/source/index.rst
@@ -14,6 +14,7 @@ PyPose Documentation
    modules
    optim
    utils
+   testing
 
 
 Indices and tables

diff --git a/docs/source/testing.rst b/docs/source/testing.rst
@@ -0,0 +1,11 @@
+Testing
+=========
+
+
+.. currentmodule:: pypose
+.. autosummary::
+   :toctree: generated
+   :nosignatures:
+
+   :template: autosummary/class-no-inherit.rst
+   testing.assert_close
diff --git a/pypose/__init__.py b/pypose/__init__.py
@@ -18,6 +18,7 @@
 from .basics import *
 from . import module
 from . import optim
+from . import testing
 
 
 min_torch = '2.0'

diff --git a/pypose/basics/ops.py b/pypose/basics/ops.py
@@ -7,7 +7,7 @@ def pm(input, *, out=None):
 
     Args:
         input (:obj:`Tensor`): the input tensor.
-    
+
     Return:
         :obj:`Tensor`: the output tensor contains only :math:`-1` or :math:`+1`.
 
@@ -103,7 +103,7 @@ def cummul(input, dim, left = True):
 
     .. math::
         y_i = x_1 * x_2 * \cdots * x_i,
-        
+
     where :math:`x_i,~y_i` are the :math:`i`-th LieType item along the :obj:`dim`
     dimension of input and output, respectively.
 
@@ -122,7 +122,7 @@ def cummul(input, dim, left = True):
           :math:`N` is the LieTensor size along the :obj:`dim` dimension.
 
     Example:
-    
+
         * Left multiplication with :math:`\text{input} \in` :obj:`SE3`
 
         >>> input = pp.randn_SE3(2)
@@ -141,7 +141,7 @@ def cummul(input, dim, left = True):
     """
     if left:
         return cumops(input, dim, lambda a, b : a * b)
-    else: 
+    else:
         return cumops(input, dim, lambda a, b : b * a)
 
 
@@ -152,7 +152,7 @@ def cumprod(input, dim, left = True):
 
     .. math::
         y_i = x_i ~\times~ x_{i-1} ~\times~ \cdots ~\times~ x_1,
-    
+
     * Right product:
 
     .. math::

diff --git a/pypose/testing/__init__.py b/pypose/testing/__init__.py
@@ -0,0 +1 @@
+from .comparison import *
diff --git a/pypose/testing/comparison.py b/pypose/testing/comparison.py
@@ -0,0 +1,42 @@
+import torch
+from ..function.checking import is_lietensor
+
+
+def assert_close(actual, expected, *args, **kwargs):
+    '''
+    Asserts that ``actual`` and ``expected`` are close. This function is exactly the same
+    with `torch.testing.assert_close <https://tinyurl.com/3bm33ps7>`_ except for that it
+    also accepts ``pypose.LieTensor``.
+
+    Args:
+        actual(:obj:`Tensor` or :obj:`LieTensor`): Actual input.
+        expected(:obj:`Tensor` or :obj:`LieTensor`): Expected input.
+        rtol (Optional[float]): Relative tolerance. If specified ``atol`` must also be
+            specified. If omitted, default values based on the :attr:`~torch.Tensor.dtype`
+            are selected with the below table.
+        atol (Optional[float]): Absolute tolerance. If specified ``rtol`` must also be
+            specified. If omitted, default values based on the :attr:`~torch.Tensor.dtype`
+            are selected with the below table.
+
+    If :math:`T_e` and :math:`T_a` are Lietensor, they are considered close if
+    :math:`T_e*T_a^{-1} = \mathbf{O}`, where :math:`\mathbf{O}` is close to zero tensor in
+    the sense of ``torch.testing.assert_close`` is ``True.``
+
+    Warning:
+        The prerequisites for the other arguments align precisely with
+        `torch.testing.assert_close <https://tinyurl.com/3bm33ps7>`_. Kindly consult it
+        for further details.
+
+    Examples:
+        >>> import pypose as pp
+        >>> actual = pp.randn_SE3(3)
+        >>> expected = actual.Log().Exp()
+        >>> pp.testing.assert_close(actual, expected, rtol=1e-5, atol=1e-5)
+
+    '''
+    if is_lietensor(actual) and is_lietensor(expected):
+        source = (actual.Inv() @ expected).Log().tensor()
+        target = torch.zeros_like(source)
+    else:
+        source, target = actual, expected
+    torch.testing.assert_close(source, target, *args, **kwargs)
diff --git a/tests/function/test_spline.py b/tests/function/test_spline.py
@@ -6,51 +6,45 @@ class TestSpline:
 
     def test_bsplilne(self):
         device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
         # test two poses
-        data = pp.randn_SE3(2, device=device)
+        data = pp.randn_SE3(2,device=device)
         poses = pp.bspline(data, 0.10, True)
-        torch.testing.assert_close(poses[...,[0,-1],:].translation(),
-                                   data[...,[0,-1],:].translation())
-        torch.testing.assert_close(poses[...,[0,-1],:].rotation(),
-                                   data[...,[0,-1],:].rotation())
+        pp.testing.assert_close(poses[...,[0,-1],:], data[...,[0,-1],:])
+
         # test for multi batch
         data = pp.randn_SE3(2,5, device=device)
         poses = pp.bspline(data, 0.5)
         assert poses.lshape[-1] == 2 * (data.lshape[-1]-3)+1
         poses = pp.bspline(data, 0.5, True)
-        torch.testing.assert_close(poses[...,[0,-1],:].translation(),
-                                   data[...,[0,-1],:].translation())
-        torch.testing.assert_close(poses[...,[0,-1],:].rotation(),
-                                   data[...,[0,-1],:].rotation())
+        pp.testing.assert_close(poses[...,[0,-1],:], data[...,[0,-1],:])
+
         # test for high dimension
-        data = pp.randn_SE3(2, 3, 4, device=device)
-        poses = pp.bspline(data, 0.20)
+        data = pp.randn_SE3(2,3,4, device=device)
+        poses = pp.bspline(data, 0.2)
         assert poses.lshape[-1] == 5 * (data.lshape[-1]-3)+1
-        poses = pp.bspline(data, 0.5, True)
-        torch.testing.assert_close(poses[...,[0,-1],:].translation(),
-                                   data[...,[0,-1],:].translation())
-        torch.testing.assert_close(poses[...,[0,-1],:].rotation(),
-                                   data[...,[0,-1],:].rotation())
+        poses = pp.bspline(data, 0.2, True)
+        pp.testing.assert_close(poses[...,[0,-1],:], data[...,[0,-1],:])
 
-        data = pp.randn_SE3(2, 3, 4, device=device)
+        data = pp.randn_SE3(2,3,4, device=device)
         poses = pp.bspline(data, 0.3)
         assert poses.lshape[-1] == 4 * (data.lshape[-1]-3)+1
-        poses = pp.bspline(data, 0.5, True)
-        torch.testing.assert_close(poses[...,[0,-1],:].translation(),
-                                   data[...,[0,-1],:].translation())
-        torch.testing.assert_close(poses[...,[0,-1],:].rotation(),
-                                   data[...,[0,-1],:].rotation())
+        poses = pp.bspline(data, 0.3, True)
+        pp.testing.assert_close(poses[...,[0,-1],:], data[...,[0,-1],:])
+
     def test_chspline(self):
         device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
         # test for different point dimension
-        points = torch.randn(1, 2, 3, device=device)
+        points = torch.randn(1,2,3, device=device)
         interval = .2
         interpoints = pp.chspline(points, interval=interval)
         num = points.shape[-2]
         k = math.ceil(1.0 / interval)
         index = k*torch.arange(0, num, device=device, dtype=torch.int64)
         po = interpoints.index_select(-2, index)
-        assert (points - po).sum() < 1e-5
+        pp.testing.assert_close(points, po)
+
         # test multi points
         points = torch.randn(20, 3, device=device)
         interval = 0.5
@@ -59,33 +53,36 @@ def test_chspline(self):
         k = math.ceil(1.0 / interval)
         index = k*torch.arange(0, num, device=device, dtype=torch.int64)
         po = interpoints.index_select(-2, index)
-        assert (points - po).sum() < 1e-5
+        pp.testing.assert_close(points, po)
+
         # test multi batches
-        points = torch.randn(3, 20, 3, device=device)
+        points = torch.randn(3,20,3, device=device)
         interval = 0.4
         interpoints = pp.chspline(points, interval=interval)
         num = points.shape[-2]
         k = math.ceil(1.0/interval)
         index = k*torch.arange(0, num, device=device, dtype=torch.int64)
         po = interpoints.index_select(-2, index)
-        assert (points - po).sum() < 1e-5
+        pp.testing.assert_close(points, po)
+
         # test multi dim of points
-        points = torch.randn(2, 3, 50, 4, device=device)
+        points = torch.randn(2,3,50,4, device=device)
         interval = 0.1
         interpoints = pp.chspline(points, interval=interval)
         num = points.shape[-2]
         k = math.ceil(1.0 / interval)
         index = k*torch.arange(0, num, device=device, dtype=torch.int64)
         po = interpoints.index_select(-2, index)
-        assert (points - po).sum() < 1e-5
-        points = torch.randn(10, 2, 3, 50, 4, device=device)
+        pp.testing.assert_close(points, po)
+
+        points = torch.randn(10,2,3,50,4, device=device)
         interval = 0.1
         interpoints = pp.chspline(points, interval=interval)
         num = points.shape[-2]
         k = math.ceil(1.0 / interval)
         index = k*torch.arange(0, num, device=device, dtype=torch.int64)
         po = interpoints.index_select(-2, index)
-        assert (points - po).sum() < 1e-5
+        pp.testing.assert_close(points, po)
 
 
 if __name__=="__main__":
-Original file line number
+Diff line change
@@ Expand Up / @@ -14,6 +14,7 @@ PyPose Documentation @@
        modules
        optim
        utils
+       testing
     Indices and tables
@@ Expand Down @@