search

craigleili / 3DLocalMultiViewDesc

End-to-End Learning Local Multi-view Descriptors for 3D Point Clouds (CVPR2020)
MIT License
60 stars 8 forks source link

Error on compile with Ninja #1

Closed Hyeokreal closed 4 years ago

Hyeokreal commented 4 years ago

Hi Lei Li,

Really thanks for sharing your wonderful works.

I have an issue on compiling the soft_renderer. When I tried to initialize MultiViewRenderer class, I got following errors.

ninja: build stopped: subcommand failed. Traceback (most recent call last): File "/usr/local/lib/python3.6/dist-packages/torch/utils/cpp_extension.py", line 1515, in _run_ninja_build env=env) File "/usr/lib/python3.6/subprocess.py", line 438, in run output=stdout, stderr=stderr) subprocess.CalledProcessError: Command '['ninja', '-v']' returned non-zero exit status 1.

During handling of the above exception, another exception occurred:

Traceback (most recent call last): File "transform.py", line 18, in from soft_renderer.cuda.jit import soft_rasterize_cuda as src File "/host_temp/codes/cqcpu/point_backends/learnable_viewpoint/soft_renderer/cuda/jit.py", line 9, in verbose=True) File "/usr/local/lib/python3.6/dist-packages/torch/utils/cpp_extension.py", line 974, in load keep_intermediates=keep_intermediates) File "/usr/local/lib/python3.6/dist-packages/torch/utils/cpp_extension.py", line 1179, in _jit_compile with_cuda=with_cuda) File "/usr/local/lib/python3.6/dist-packages/torch/utils/cpp_extension.py", line 1279, in _write_ninja_file_and_build_library error_prefix="Error building extension '{}'".format(name)) File "/usr/local/lib/python3.6/dist-packages/torch/utils/cpp_extension.py", line 1529, in _run_ninja_build raise RuntimeError(message) RuntimeError: Error building extension 'soft_rasterize_cuda'

I tried within both my local machine and a docker container that you provided but in both environment same errors occured.

Do you have any idea on this issue ?

Thanks.

Hyeokreal commented 4 years ago

It was a pytorch version issue. Thanks

chrisranderson commented 3 years ago

@Hyeokreal What version worked for you?

AndreiMiculita commented 3 years ago

This might be useful to other people coming across this. I also couldn't compile the soft_renderer at first.

libcudart.so was not being found even though I had installed it as per the instructions on nVidia's website. What fixed it was just installing libtorch3-dev via apt.

For me it worked with torch==1.2.0 (as in requirements.txt, this was not an issue).

Edit: that was probably not it. It stopped working today for what I can only assume is some nondeterministic reason. I got it working again by switching to a more recent torch version, such that the CUDA version torch is built with matches the version of CUDA installed on the system.