"RuntimeError: Not implemented on the CPU." in deform_source_mesh_to_target_mesh.ipynb

[semihcanturk]

When running the deform_source_mesh_to_target_mesh.ipynb tutorial, I encounter the following error in the 'sample_points_from_meshes()' function in the main optimization loop, cell 10 under section "3. Optimization loop":

RuntimeError                              Traceback (most recent call last)

<ipython-input-10-eaad4829c9f7> in <module>
     28 
     29     # We sample 5k points from the surface of each mesh
---> 30     sample_trg = sample_points_from_meshes(trg_mesh, 5000)
     31     sample_src = sample_points_from_meshes(new_src_mesh, 5000)
     32 

~/PycharmProjects/pytorch3d/pytorch3d/ops/sample_points_from_meshes.py in sample_points_from_meshes(meshes, num_samples, return_normals)
     53     with torch.no_grad():
     54         areas, _ = _C.face_areas_normals(
---> 55             verts, faces
     56         )  # Face areas can be zero.
     57         max_faces = meshes.num_faces_per_mesh().max().item()

RuntimeError: Not implemented on the CPU. (face_areas_normals at /Users/semo/PycharmProjects/pytorch3d/pytorch3d/csrc/face_areas_normals/face_areas_normals.h:35)
frame #0: c10::Error::Error(c10::SourceLocation, std::__1::basic_string<char, std::__1::char_traits<char>, std::__1::allocator<char> > const&) + 135 (0x12737c9e7 in libc10.dylib)
frame #1: face_areas_normals(at::Tensor, at::Tensor) + 857 (0x127e8e679 in _C.cpython-37m-darwin.so)
frame #2: std::__1::tuple<at::Tensor, at::Tensor> pybind11::detail::argument_loader<at::Tensor, at::Tensor>::call_impl<std::__1::tuple<at::Tensor, at::Tensor>, std::__1::tuple<at::Tensor, at::Tensor> (*&)(at::Tensor, at::Tensor), 0ul, 1ul, pybind11::detail::void_type>(std::__1::tuple<at::Tensor, at::Tensor> (*&)(at::Tensor, at::Tensor), std::__1::integer_sequence<unsigned long, 0ul, 1ul>, pybind11::detail::void_type&&) + 57 (0x127e96fa9 in _C.cpython-37m-darwin.so)
frame #3: void pybind11::cpp_function::initialize<std::__1::tuple<at::Tensor, at::Tensor> (*&)(at::Tensor, at::Tensor), std::__1::tuple<at::Tensor, at::Tensor>, at::Tensor, at::Tensor, pybind11::name, pybind11::scope, pybind11::sibling>(std::__1::tuple<at::Tensor, at::Tensor> (*&)(at::Tensor, at::Tensor), std::__1::tuple<at::Tensor, at::Tensor> (*)(at::Tensor, at::Tensor), pybind11::name const&, pybind11::scope const&, pybind11::sibling const&)::'lambda'(pybind11::detail::function_call&)::operator()(pybind11::detail::function_call&) const + 135 (0x127e96857 in _C.cpython-37m-darwin.so)
frame #4: pybind11::cpp_function::dispatcher(_object*, _object*, _object*) + 3088 (0x127e98350 in _C.cpython-37m-darwin.so)
frame #5: _PyMethodDef_RawFastCallKeywords + 537 (0x108117865 in Python)
frame #6: _PyCFunction_FastCallKeywords + 41 (0x108116dc0 in Python)
frame #7: call_function + 628 (0x1081ac2b4 in Python)
frame #8: _PyEval_EvalFrameDefault + 6767 (0x1081a528f in Python)
frame #9: _PyEval_EvalCodeWithName + 1698 (0x1081acb15 in Python)
frame #10: _PyFunction_FastCallKeywords + 212 (0x108116d88 in Python)
frame #11: call_function + 737 (0x1081ac321 in Python)
frame #12: _PyEval_EvalFrameDefault + 6922 (0x1081a532a in Python)
frame #13: _PyEval_EvalCodeWithName + 1698 (0x1081acb15 in Python)
frame #14: PyEval_EvalCode + 51 (0x1081a377d in Python)
frame #15: builtin_exec + 563 (0x1081a1207 in Python)
frame #16: _PyMethodDef_RawFastCallKeywords + 488 (0x108117834 in Python)
frame #17: _PyCFunction_FastCallKeywords + 41 (0x108116dc0 in Python)
frame #18: call_function + 628 (0x1081ac2b4 in Python)
frame #19: _PyEval_EvalFrameDefault + 6922 (0x1081a532a in Python)
frame #20: gen_send_ex + 244 (0x1081227e7 in Python)
frame #21: _PyEval_EvalFrameDefault + 17461 (0x1081a7c55 in Python)
frame #22: gen_send_ex + 244 (0x1081227e7 in Python)
frame #23: _PyEval_EvalFrameDefault + 17461 (0x1081a7c55 in Python)
frame #24: gen_send_ex + 244 (0x1081227e7 in Python)
frame #25: _PyMethodDef_RawFastCallKeywords + 583 (0x108117893 in Python)
frame #26: _PyMethodDescr_FastCallKeywords + 81 (0x10811c065 in Python)
frame #27: call_function + 782 (0x1081ac34e in Python)
frame #28: _PyEval_EvalFrameDefault + 6742 (0x1081a5276 in Python)
frame #29: function_code_fastcall + 106 (0x108117194 in Python)
frame #30: call_function + 737 (0x1081ac321 in Python)
frame #31: _PyEval_EvalFrameDefault + 6922 (0x1081a532a in Python)
frame #32: function_code_fastcall + 106 (0x108117194 in Python)
frame #33: call_function + 737 (0x1081ac321 in Python)
frame #34: _PyEval_EvalFrameDefault + 6742 (0x1081a5276 in Python)
frame #35: _PyEval_EvalCodeWithName + 1698 (0x1081acb15 in Python)
frame #36: _PyFunction_FastCallDict + 444 (0x108116a00 in Python)
frame #37: _PyObject_Call_Prepend + 131 (0x108117b07 in Python)
frame #38: PyObject_Call + 136 (0x108116ecd in Python)
frame #39: _PyEval_EvalFrameDefault + 7507 (0x1081a5573 in Python)
frame #40: _PyEval_EvalCodeWithName + 1698 (0x1081acb15 in Python)
frame #41: _PyFunction_FastCallKeywords + 212 (0x108116d88 in Python)
frame #42: call_function + 737 (0x1081ac321 in Python)
frame #43: _PyEval_EvalFrameDefault + 7090 (0x1081a53d2 in Python)
frame #44: gen_send_ex + 244 (0x1081227e7 in Python)
frame #45: builtin_next + 99 (0x1081a1cfe in Python)
frame #46: _PyMethodDef_RawFastCallKeywords + 488 (0x108117834 in Python)
frame #47: _PyCFunction_FastCallKeywords + 41 (0x108116dc0 in Python)
frame #48: call_function + 628 (0x1081ac2b4 in Python)
frame #49: _PyEval_EvalFrameDefault + 6922 (0x1081a532a in Python)
frame #50: _PyEval_EvalCodeWithName + 1698 (0x1081acb15 in Python)
frame #51: _PyFunction_FastCallKeywords + 212 (0x108116d88 in Python)
frame #52: call_function + 737 (0x1081ac321 in Python)
frame #53: _PyEval_EvalFrameDefault + 6742 (0x1081a5276 in Python)
frame #54: gen_send_ex + 244 (0x1081227e7 in Python)
frame #55: builtin_next + 99 (0x1081a1cfe in Python)
frame #56: _PyMethodDef_RawFastCallKeywords + 488 (0x108117834 in Python)
frame #57: _PyCFunction_FastCallKeywords + 41 (0x108116dc0 in Python)
frame #58: call_function + 628 (0x1081ac2b4 in Python)
frame #59: _PyEval_EvalFrameDefault + 6922 (0x1081a532a in Python)
frame #60: _PyEval_EvalCodeWithName + 1698 (0x1081acb15 in Python)
frame #61: _PyFunction_FastCallKeywords + 212 (0x108116d88 in Python)
frame #62: call_function + 737 (0x1081ac321 in Python)
frame #63: _PyEval_EvalFrameDefault + 6922 (0x1081a532a in Python)

Instructions To Reproduce the Issue:

1. Clone & build the latest master (v 0.1)
2. Run the deform_source_mesh_to_target_mesh.ipynb notebook on the CPU as opposed to CUDA. Make sure to have the example target 3D model downloaded and placed in the appropriate directory as instructed in cell 3.

[nikhilaravi]

@semihcanturk thanks for raising this issue! We are currently working on more complete CPU support for all CUDA operators (face_areas_normals only has a CUDA implementation). In the meantime can you try running this on a GPU with Colab?

[nikhilaravi]

@semihcanturk this has now been fixed on master - I tested this tutorial on cpu on colab and it runs without errors (although note that the optimization will run more slowly).