remove C++ for portability/compatibility

Author: Peter

pyproject.toml

@@ -71,5 +71,5 @@ mujoco = ["mujoco"]
 [build-system]
 # Including torch and ninja here are needed to build the native code.
 # They will be installed as dependencies during the build, which can take a while the first time.
-requires = ["setuptools>=60.0.0", "wheel", "torch==2.1.0", "ninja"]
+requires = ["setuptools>=60.0.0", "wheel"]
 build-backend= "setuptools.build_meta"

setup.py

@@ -281,44 +281,7 @@ def get_link_names(self):
     def get_frame_indices(self, *frame_names):
         return torch.tensor([self.frame_to_idx[n] for n in frame_names], dtype=torch.long,
                             device=self.device)
-
     def forward_kinematics(self, th, frame_indices: Optional = None):
-        """
-        Compute forward kinematics.
-
-        Args:
-            th: A dict, list, numpy array, or torch tensor of ALL joints values. Possibly batched.
-               the fastest thing to use is a torch tensor, all other types get converted to that.
-               If any joint values are missing, an exception will be thrown.
-            frame_indices: A list of frame indices to compute forward kinematics for. If None, all frames are computed.
-
-        Returns:
-            A dict of frame names and their corresponding Transform3d objects.
-        """
-        if frame_indices is None:
-            frame_indices = self.get_all_frame_indices()
-
-        th = self.ensure_tensor(th)
-        th = torch.atleast_2d(th)
-
-        import zpk_cpp
-        frame_transforms = zpk_cpp.fk(
-            frame_indices,
-            self.axes,
-            th,
-            self.parents_indices,
-            self.joint_type_indices,
-            self.joint_indices,
-            self.joint_offsets,
-            self.link_offsets
-        )
-
-        frame_names_and_transform3ds = {self.idx_to_frame[frame_idx]: tf.Transform3d(matrix=transform) for
-                                        frame_idx, transform in frame_transforms.items()}
-
-        return frame_names_and_transform3ds
-
-    def forward_kinematics_py(self, th, frame_indices: Optional = None):
         if frame_indices is None:
             frame_indices = self.get_all_frame_indices()
 

src/pytorch_kinematics/chain.py

@@ -255,34 +255,9 @@ def test_mjcf_slide_joint_parsing():
 
 
 def test_fk_val():
-    dtype = torch.float64
-    d = "cuda" if torch.cuda.is_available() else "cpu"
-
     chain = pk.build_chain_from_mjcf(open(os.path.join(TEST_DIR, "val.xml")).read())
-    chain = chain.to(dtype=torch.float64, device=d)
-
-    th = torch.rand(1000, chain.n_joints, dtype=dtype, device=d)
-
-    def _fk_no_compile():
-        return chain.forward_kinematics_py(th)
-
-    @torch.compile(backend='inductor')
-    def _fk_compile():
-        return chain.forward_kinematics_py(th)
-
-    from timeit import timeit
-
-    # warmup
-    _fk_no_compile()
-    _fk_compile()
-
-    number = 10
-    ms_no_compile = timeit(_fk_no_compile, number=number) / number * 1000
-    print(f"elapsed {ms_no_compile:.1f}ms for no compile")
-    ms_compile = timeit(_fk_compile, number=number) / number * 1000
-    print(f"elapsed {ms_compile:.1f}ms for compile")
-
-    ret = chain.forward_kinematics_py(th)
+    chain = chain.to(dtype=torch.float64)
+    ret = chain.forward_kinematics(torch.zeros([1000, chain.n_joints], dtype=torch.float64))
     tg = ret['drive45']
     pos, rot = quat_pos_from_transform3d(tg)
     torch.set_printoptions(precision=6, sci_mode=False)

src/zpk_cpp/pk.cpp

@@ -4,7 +4,6 @@
 
 from pytorch_kinematics.transforms.rotation_conversions import axis_and_angle_to_matrix, axis_angle_to_matrix, \
     pos_rot_to_matrix, matrix_to_pos_rot, random_rotations
-import zpk_cpp
 
 
 def test_axis_angle_to_matrix_perf():