Build problem about gtsam

[JIANG-CX]

When I run the command "cmake --build build_gtsam --config RelWithDebInfo -j", it gave an error as follows. Could you please help me to fix it? Thanks.

Traceback (most recent call last): File "/home/jiangcx/Documents/Nerf_related/NeRF-SLAM/thirdparty/gtsam/wrap/scripts/pybind_wrap.py", line 94, in <module> main() File "/home/jiangcx/Documents/Nerf_related/NeRF-SLAM/thirdparty/gtsam/wrap/scripts/pybind_wrap.py", line 90, in main wrapper.wrap(sources, args.out) File "/home/jiangcx/Documents/Nerf_related/NeRF-SLAM/thirdparty/gtsam/wrap/gtwrap/pybind_wrapper.py", line 723, in wrap submodules=submodules) File "/home/jiangcx/Documents/Nerf_related/NeRF-SLAM/thirdparty/gtsam/wrap/gtwrap/pybind_wrapper.py", line 624, in wrap_file module = parser.Module.parseString(content) File "/home/jiangcx/Documents/Nerf_related/NeRF-SLAM/thirdparty/gtsam/wrap/gtwrap/interface_parser/module.py", line 56, in parseString return Module.rule.parseString(s)[0] File "/opt/ros/noetic/lib/python3/dist-packages/pyparsing/core.py", line 1141, in parse_string raise exc.with_traceback(None) pyparsing.exceptions.ParseException: Expected string_end, found 'namespace' (at char 1249), (line:46, col:1) make[2]: *** [python/CMakeFiles/pybind_wrap_gtsam_unstable.dir/build.make:76: python/gtsam_unstable.cpp] Error 1 make[1]: *** [CMakeFiles/Makefile2:32620: python/CMakeFiles/pybind_wrap_gtsam_unstable.dir/all] Error 2 make[1]: *** Waiting for unfinished jobs....

[pgaston]

Same issue here. New Ubuntu 20.04 install.

[hjxwhy]

@phamquandung I have done as you said, but I still have the problem. Which version pyparsing you used?

[JIANG-CX]

@JIANG-CX, @pgaston I found the solution. Please go to the folder thirdparty and delete gtsam then git clone the updated (https://github.com/ToniRV/gtsam-1/tree/feature/nerf_slam). Then build again.

@ToniRV . I think you should update the gtsam in thirdparty to avoid this issue.

After downloading the new gtsam, I find this error still exists.

[phamquandung]

@JIANG-CX @pgaston @hjxwhy you can fork https://github.com/borglab/gtsam this one and build python wrapper like this https://github.com/borglab/gtsam/tree/develop/python. I tried, and it works.

[pgaston]

@phamquandung
Okay, still hung up here. I moved the gtsam you specified above and built it per instructions. However... when I run I get:

File "/home/pg/repos/NeRF-SLAM/./examples/../slam/vio_slam.py", line 65, in initial_state
naive_pose = gtsam.Pose3.identity()
AttributeError: type object 'gtsam.gtsam.Pose3' has no attribute 'identity'

Thoughts?

[richard7714]

@phamquandung
Okay, still hung up here. I moved the gtsam you specified above and built it per instructions. However... when I run I get:

File "/home/pg/repos/NeRF-SLAM/./examples/../slam/vio_slam.py", line 65, in initial_state
naive_pose = gtsam.Pose3.identity()
AttributeError: type object 'gtsam.gtsam.Pose3' has no attribute 'identity'

Thoughts?

I think it's a problem of gtsam version. According to borglab/gtsam#1298, identity() was replaced to Identity() (UpperCase)

[phamquandung]

Then, another error:

File "/home/user/research/NeRF-SLAM/./examples/../slam/visual_frontends/visual_frontend.py", line 48, in gtsam_pose_to_torch
q = pose.rotation().quaternion()
AttributeError: 'gtsam.gtsam.Rot3' object has no attribute 'quaternion'

[richard7714]

@phamquandung
Then, another error:

File "/home/user/research/NeRF-SLAM/./examples/../slam/visual_frontends/visual_frontend.py", line 48, in gtsam_pose_to_torch q = pose.rotation().quaternion() AttributeError: 'gtsam.gtsam.Rot3' object has no attribute 'quaternion'

That line and the following return statement should be replaced like this

    q = pose.rotation().toQuaternion()
    return torch.tensor([t[0], t[1], t[2], q.x(), q.y(), q.z(), q.w()], device=device, dtype=dtype)

[JIANG-CX]

@JIANG-CX @pgaston @hjxwhy you can fork https://github.com/borglab/gtsam this one and build python wrapper like this https://github.com/borglab/gtsam/tree/develop/python. I tried, and it works.

Thanks for your kindly help. I can pass the build step now. But when I run the "./examples/slam_demo.py”, I find there is an error about the gtsam. I wonder have you met such error? How did you solve it? Thanks a lot.
File "./examples/slam_demo.py", line 200, in <module> run(args) File "./examples/slam_demo.py", line 143, in run slam_module.spin() # visualizer should be the main spin, but pytorch has a memory bug/leak if threaded... File "./examples/../pipeline/pipeline_module.py", line 88, in spin self.initialize_module() File "./examples/../slam/slam_module.py", line 18, in initialize_module from slam.vio_slam import VioSLAM File "./examples/../slam/vio_slam.py", line 9, in <module> from factor_graph.factor_graph import TorchFactorGraph File "./examples/../factor_graph/factor_graph.py", line 14, in <module> from gtsam import NonlinearFactorGraph as FactorGraph ModuleNotFoundError: No module named 'gtsam'

[phamquandung]

@JIANG-CX @pgaston @hjxwhy you can fork https://github.com/borglab/gtsam this one and build python wrapper like this https://github.com/borglab/gtsam/tree/develop/python. I tried, and it works.

Thanks for your kindly help. I can pass the build step now. But when I run the "./examples/slam_demo.py”, I find there is an error about the gtsam. I wonder have you met such error? How did you solve it? Thanks a lot. File "./examples/slam_demo.py", line 200, in <module> run(args) File "./examples/slam_demo.py", line 143, in run slam_module.spin() # visualizer should be the main spin, but pytorch has a memory bug/leak if threaded... File "./examples/../pipeline/pipeline_module.py", line 88, in spin self.initialize_module() File "./examples/../slam/slam_module.py", line 18, in initialize_module from slam.vio_slam import VioSLAM File "./examples/../slam/vio_slam.py", line 9, in <module> from factor_graph.factor_graph import TorchFactorGraph File "./examples/../factor_graph/factor_graph.py", line 14, in <module> from gtsam import NonlinearFactorGraph as FactorGraph ModuleNotFoundError: No module named 'gtsam'

In the build folder you should run make python-install and make python-test, then it should be OK.

[JIANG-CX]

Yes, I have run make python-install and make python-test, and got the feedback "[100%] Built target python-test" and "Successfully installed gtsam-4.2a8". But the error still exits.

[jrpowers]

Maybe you installed gtsam to another python environment?

Also, I made a bunch the changes to adopt to new gtsam in this PR, in case it helps: #12

[sxyu]

Using the branch feature/nerf_slam https://github.com/ToniRV/gtsam-1/tree/feature/nerf_slam
I found that there is a syntax error in gtsam_unstable/gtsam_unstable.i
causing a parsing error:
virtual class LinearizedHessianFactor : gtsam::LinearizedGaussianFactor {
Needs to be closed 2 lines below (line 818) with };.

EDIT: the author has fixed it and the feature/nerf_slam branch of gtsam should be ok now.

[yzslab]

Maybe you installed gtsam to another python environment?

Also, I made a bunch the changes to adopt to new gtsam in this PR, in case it helps: #12

It works well! Nice job!

[pgaston]

@jrpowers Using your latest PR I got everything to build - whew. The cube sample scene runs fine! Unfortunately, the first example dies a horrible death... (perhaps I'll wait for the dust to settle on all these changes...) Thanks again to all!

(nerf) pg@pg-MS-7A59:~/repos/NeRF-SLAM$ python ./examples/slam_demo.py --dataset_dir=./datasets/Replica/office0 --dataset_name=nerf --buffer=100 --slam --img_stride=2 --fusion='nerf' --multi_gpu --gui
Running with GPUs: 0
Running pipeline in sequential mode.
Loading 1000 images.
  0%|                                                  | 0/1000 [00:00<?, ?it/s]/home/pg/miniconda3/envs/nerf/lib/python3.9/multiprocessing/resource_tracker.py:216: UserWarning: resource_tracker: There appear to be 1 leaked semaphore objects to clean up at shutdown
  warnings.warn('resource_tracker: There appear to be %d '
Segmentation fault (core dumped)

The "GPUs: 0" looks fishy. However, nvidia-smi; nvcc --version; and nvitop -- all seem to be showing the usual results.

[sxyu]

works fully for me with the fixed GTSAM from the author & (only) python changes in this repo made by jrpowers

[pengwangucla]

what is you gcc version or other environment setup ? , after switch the branch. met this error

/usr/include/c++/8/bits/alloc_traits.h:392:27: error: forming pointer to reference type ‘const Eigen::Matrix<double, -1, -1>&’
       using pointer = _Tp*;
                           ^
/usr/include/c++/8/bits/alloc_traits.h:395:39: error: forming pointer to reference type ‘const Eigen::Matrix<double, -1, -1>&’
       using const_pointer = const _Tp*;

Solved by checking:

works fully for me with the fixed GTSAM from the author & (only) python changes in this repo made by jrpowers

[JIANG-CX]

what is you gcc version or other environment setup ? , after switch the branch. met this error

/usr/include/c++/8/bits/alloc_traits.h:392:27: error: forming pointer to reference type ‘const Eigen::Matrix<double, -1, -1>&’
       using pointer = _Tp*;
                           ^
/usr/include/c++/8/bits/alloc_traits.h:395:39: error: forming pointer to reference type ‘const Eigen::Matrix<double, -1, -1>&’
       using const_pointer = const _Tp*;

Solved by checking:

works fully for me with the fixed GTSAM from the author & (only) python changes in this repo made by jrpowers

I do not understand this comment. Could you tell me how to fix the linear.cpp? What's the "inner const X& ..." stands for?

[OceanYing]

what is you gcc version or other environment setup ? , after switch the branch. met this error

/usr/include/c++/8/bits/alloc_traits.h:392:27: error: forming pointer to reference type ‘const Eigen::Matrix<double, -1, -1>&’
       using pointer = _Tp*;
                           ^
/usr/include/c++/8/bits/alloc_traits.h:395:39: error: forming pointer to reference type ‘const Eigen::Matrix<double, -1, -1>&’
       using const_pointer = const _Tp*;

Solved by checking:

works fully for me with the fixed GTSAM from the author & (only) python changes in this repo made by jrpowers

I do not understand this comment. Could you tell me how to fix the linear.cpp? What's the "inner const X& ..." stands for?

It seems that the solution comes from README installation-issues: 3. But i'm also not clear about the "inner const X& ..." and still stuck with this problem now. If you have figured it out, please let me know. @JIANG-CX @pengwangucla

[pengwangucla]

You may change const std::vector<const gtsam::Matrix&>& to const std::vector<gtsam::Matrix>& use & as pointer is not proper define of std::vector

[OceanYing]

You may change const std::vector<const gtsam::Matrix&>& to const std::vector<gtsam::Matrix>& use & as pointer is not proper define of std::vector

Thanks a lot! That's the solution and it works now!

[yusuf-wadi]

@JIANG-CX, @pgaston I found the solution. Please go to the folder thirdparty and delete gtsam then git clone the updated (https://github.com/ToniRV/gtsam-1/tree/feature/nerf_slam). Then build again.
@ToniRV . I think you should update the gtsam in thirdparty to avoid this issue.

After downloading the new gtsam, I find this error still exists.

this worked for me

[KnightOfTheMoonlight]

@phamquandung
Then, another error:
File "/home/user/research/NeRF-SLAM/./examples/../slam/visual_frontends/visual_frontend.py", line 48, in gtsam_pose_to_torch q = pose.rotation().quaternion() AttributeError: 'gtsam.gtsam.Rot3' object has no attribute 'quaternion'

That line and the following return statement should be replaced like this

    q = pose.rotation().toQuaternion()
    return torch.tensor([t[0], t[1], t[2], q.x(), q.y(), q.z(), q.w()], device=device, dtype=dtype)

Thanks for your reply! It saved my day! I have run into an error because of wrong modification of this.

My wrong version is pose.rotation().toQuaternion().coeffs();
This modification leads q to all nan

[Tonny24Wang]

When I run the command "cmake --build build_gtsam --config RelWithDebInfo -j", it gave an error as follows. Could you please help me to fix it? Thanks.

Traceback (most recent call last): File "/home/jiangcx/Documents/Nerf_related/NeRF-SLAM/thirdparty/gtsam/wrap/scripts/pybind_wrap.py", line 94, in <module> main() File "/home/jiangcx/Documents/Nerf_related/NeRF-SLAM/thirdparty/gtsam/wrap/scripts/pybind_wrap.py", line 90, in main wrapper.wrap(sources, args.out) File "/home/jiangcx/Documents/Nerf_related/NeRF-SLAM/thirdparty/gtsam/wrap/gtwrap/pybind_wrapper.py", line 723, in wrap submodules=submodules) File "/home/jiangcx/Documents/Nerf_related/NeRF-SLAM/thirdparty/gtsam/wrap/gtwrap/pybind_wrapper.py", line 624, in wrap_file module = parser.Module.parseString(content) File "/home/jiangcx/Documents/Nerf_related/NeRF-SLAM/thirdparty/gtsam/wrap/gtwrap/interface_parser/module.py", line 56, in parseString return Module.rule.parseString(s)[0] File "/opt/ros/noetic/lib/python3/dist-packages/pyparsing/core.py", line 1141, in parse_string raise exc.with_traceback(None) pyparsing.exceptions.ParseException: Expected string_end, found 'namespace' (at char 1249), (line:46, col:1) make[2]: *** [python/CMakeFiles/pybind_wrap_gtsam_unstable.dir/build.make:76: python/gtsam_unstable.cpp] Error 1 make[1]: *** [CMakeFiles/Makefile2:32620: python/CMakeFiles/pybind_wrap_gtsam_unstable.dir/all] Error 2 make[1]: *** Waiting for unfinished jobs....

thirdparty/gtsam/gtsam_unstable/gtsam_unstable.i
815,1: virtual class LinearizedHessianFactor : gtsam::LinearizedGaussianFactor {
这个类结尾 少了两个字符 }; ，加上即可

[SimonsRoad]

try "git clone https://github.com/ToniRV/gtsam-1"
then cmake it!

[lllray]

try "git clone https://github.com/ToniRV/gtsam-1" then cmake it!

this worked for me.

[wbjsamuel]

@jrpowers Using your latest PR I got everything to build - whew. The cube sample scene runs fine! Unfortunately, the first example dies a horrible death... (perhaps I'll wait for the dust to settle on all these changes...) Thanks again to all!

(nerf) pg@pg-MS-7A59:~/repos/NeRF-SLAM$ python ./examples/slam_demo.py --dataset_dir=./datasets/Replica/office0 --dataset_name=nerf --buffer=100 --slam --img_stride=2 --fusion='nerf' --multi_gpu --gui
Running with GPUs: 0
Running pipeline in sequential mode.
Loading 1000 images.
  0%|                                                  | 0/1000 [00:00<?, ?it/s]/home/pg/miniconda3/envs/nerf/lib/python3.9/multiprocessing/resource_tracker.py:216: UserWarning: resource_tracker: There appear to be 1 leaked semaphore objects to clean up at shutdown
  warnings.warn('resource_tracker: There appear to be %d '
Segmentation fault (core dumped)

The "GPUs: 0" looks fishy. However, nvidia-smi; nvcc --version; and nvitop -- all seem to be showing the usual results.

Hi I also met this segmentation fault error. Could you please tell me how did you solve this issue? Thx a lot!

[Zhangxr02]

@JIANG-CX @pgaston @hjxwhy you can fork https://github.com/borglab/gtsam this one and build python wrapper like this https://github.com/borglab/gtsam/tree/develop/python. I tried, and it works.

Thanks for your kindly help. I can pass the build step now. But when I run the "./examples/slam_demo.py”, I find there is an error about the gtsam. I wonder have you met such error? How did you solve it? Thanks a lot. File "./examples/slam_demo.py", line 200, in <module> run(args) File "./examples/slam_demo.py", line 143, in run slam_module.spin() # visualizer should be the main spin, but pytorch has a memory bug/leak if threaded... File "./examples/../pipeline/pipeline_module.py", line 88, in spin self.initialize_module() File "./examples/../slam/slam_module.py", line 18, in initialize_module from slam.vio_slam import VioSLAM File "./examples/../slam/vio_slam.py", line 9, in <module> from factor_graph.factor_graph import TorchFactorGraph File "./examples/../factor_graph/factor_graph.py", line 14, in <module> from gtsam import NonlinearFactorGraph as FactorGraph ModuleNotFoundError: No module named 'gtsam'

Hello, I have the same problem when I run the demo. Although I have update gtsam and torch118, it didn't work. So how did you solve the problem(No module named 'gtsam')?