FAILED: depthconv_cuda_redo.o

[kuzhang]

Evaluating 1 trackers on 15 sequences
Tracker: dimp_rgbd_blend dimp50_votd None , Sequence: multi/multi_22
have 1672 depth frames
['loading net --', 'dimp50.pth']
Song in pytracking.utils.loading.py load_network ..
Song in ltr.admin.loading.py: line 65, net_constr.kwds.keys() arg = backbone_pretrained
Song in ltr.admin.model_constructor: fun_module : ltr.models.tracking.dimpnet
Song in ltr.admin.model_constructor: fun_name : dimpnet50
Using /home/dlp/.cache/torch_extensions/py37_cu102 as PyTorch extensions root...
Detected CUDA files, patching ldflags
Emitting ninja build file /home/dlp/.cache/torch_extensions/py37_cu102/_depthconv/build.ninja...
Building extension module _depthconv...
Allowing ninja to set a default number of workers... (overridable by setting the environment variable MAX_JOBS=N)
[1/2] c++ -MMD -MF depthconv_cuda_redo.o.d -DTORCH_EXTENSION_NAME=_depthconv -DTORCH_API_INCLUDE_EXTENSION_H -DPYBIND11_COMPILER_TYPE="_gcc" -DPYBIND11_STDLIB="_libstdcpp" -DPYBIND11_BUILD_ABI="_cxxabi1011" -isystem /home/dlp/anaconda3/envs/dal/lib/python3.7/site-packages/torch/include -isystem /home/dlp/anaconda3/envs/dal/lib/python3.7/site-packages/torch/include/torch/csrc/api/include -isystem /home/dlp/anaconda3/envs/dal/lib/python3.7/site-packages/torch/include/TH -isystem /home/dlp/anaconda3/envs/dal/lib/python3.7/site-packages/torch/include/THC -isystem /home/dlp/anaconda3/envs/dal/include/python3.7m -D_GLIBCXX_USE_CXX11_ABI=0 -fPIC -std=c++14 -c /home/dlp/github/DAL/pytracking_dimp/ltr/external/depthconv/src_pytorch13/depthconv_cuda_redo.c -o depthconv_cuda_redo.o
FAILED: depthconv_cuda_redo.o
c++ -MMD -MF depthconv_cuda_redo.o.d -DTORCH_EXTENSION_NAME=_depthconv -DTORCH_API_INCLUDE_EXTENSION_H -DPYBIND11_COMPILER_TYPE="_gcc" -DPYBIND11_STDLIB="_libstdcpp" -DPYBIND11_BUILD_ABI="_cxxabi1011" -isystem /home/dlp/anaconda3/envs/dal/lib/python3.7/site-packages/torch/include -isystem /home/dlp/anaconda3/envs/dal/lib/python3.7/site-packages/torch/include/torch/csrc/api/include -isystem /home/dlp/anaconda3/envs/dal/lib/python3.7/site-packages/torch/include/TH -isystem /home/dlp/anaconda3/envs/dal/lib/python3.7/site-packages/torch/include/THC -isystem /home/dlp/anaconda3/envs/dal/include/python3.7m -D_GLIBCXX_USE_CXX11_ABI=0 -fPIC -std=c++14 -c /home/dlp/github/DAL/pytracking_dimp/ltr/external/depthconv/src_pytorch13/depthconv_cuda_redo.c -o depthconv_cuda_redo.o
/home/dlp/github/DAL/pytracking_dimp/ltr/external/depthconv/src_pytorch13/depthconv_cuda_redo.c:1:10: fatal error: THC/THC.h: No such file or directory
1 | #include <THC/THC.h>
| ^~~~~~~~~~~
compilation terminated.
ninja: build stopped: subcommand failed.
Error building extension '_depthconv'

[kuzhang]

Hi there, I tried to use your code to test on custom RGBD data, I had set up the env as pytracking and I got the above error while doing the evaluation. It refers to ninja and depth_conv error. DO you have any idea why this error occurs?

[xiaozai]

Hi, sorry for the late reply, just came back from the holidays. To be honest, I met this error too and I can not compile DepthConv sometimes too, it is because of the version of PyTorch. The original depthconv is too old and it will raise the compile error when you use a newer PyTorch. You may check the DepthConv and I found other people also met similar errors.