dcn编译失败（Compile dcn dailed）

[prz30]

我在编译dcn的时候遇到了错误，显示deform_conv_cuda.o编译失败，请问有人遇到相同的问题吗？
when i compile dcn， i meet a failed which tell the deform_conv_cuda.o is fail to compile
running build_ext
building 'deform_conv_cuda' extension
Emitting ninja build file /workspace/DB-master/assets/ops/dcn/build/temp.linux-x86_64-3.6/build.ninja...
Compiling objects...
Allowing ninja to set a default number of workers... (overridable by setting the environment variable MAX_JOBS=N)
[1/1] c++ -MMD -MF /workspace/DB-master/assets/ops/dcn/build/temp.linux-x86_64-3.6/src/deform_conv_cuda.o.d -pthread -B /home/work/.conda/envs/DB/compiler_compat -Wl,--sysroot=/ -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -fPIC -I/home/work/.conda/envs/DB/lib/python3.6/site-packages/torch/include -I/home/work/.conda/envs/DB/lib/python3.6/site-packages/torch/include/torch/csrc/api/include -I/home/work/.conda/envs/DB/lib/python3.6/site-packages/torch/include/TH -I/home/work/.conda/envs/DB/lib/python3.6/site-packages/torch/include/THC -I/usr/local/cuda/include -I/home/work/.conda/envs/DB/include/python3.6m -c -c /workspace/DB-master/assets/ops/dcn/src/deform_conv_cuda.cpp -o /workspace/DB-master/assets/ops/dcn/build/temp.linux-x86_64-3.6/src/deform_conv_cuda.o -DTORCH_API_INCLUDE_EXTENSION_H -DTORCH_EXTENSION_NAME=deform_conv_cuda -D_GLIBCXX_USE_CXX11_ABI=0 -std=c++14
FAILED: /workspace/DB-master/assets/ops/dcn/build/temp.linux-x86_64-3.6/src/deform_conv_cuda.o
/workspace/DB-master/assets/ops/dcn/src/deform_conv_cuda.cpp: In function 'int deform_conv_forward_cuda(at::Tensor, at::Tensor, at::Tensor, at::Tensor, at::Tensor, at::Tensor, int, int, int, int, int, int, int, int, int, int, int)':
/workspace/DB-master/assets/ops/dcn/src/deform_conv_cuda.cpp:192:3: error: 'AT_CHECK' was not declared in this scope
AT_CHECK((offset.size(0) == batchSize), "invalid batch size of offset");
^~~~~~~~

[longnanxi]

你好，请问这个问题解决了吗？

[prz30]

解决了，这个问题是因为pytorch1.3之后的版本c++ 扩展那块用TORCH_CHECK替换掉了AT_CHECK，在deform_conv_cuda.cpp和deform_pool_cuda.cpp中将相应函数换掉就可以了

[allenmatt98]

Solved, this problem is because the version of C++ after pytorch1.3 has replaced AT_CHECK with TORCH_CHECK in the version of pytorch1.3 and then replaced the corresponding function in deform_conv_cuda.cpp and deform_pool_cuda.cpp.

Hey I'm facing a similar issue when I run python setup.py build_ext --inplace in dcn. Could you explain in depth how to fix this issue

[prz30]

Solved, this problem is because the version of C++ after pytorch1.3 has replaced AT_CHECK with TORCH_CHECK in the version of pytorch1.3 and then replaced the corresponding function in deform_conv_cuda.cpp and deform_pool_cuda.cpp.

Hey I'm facing a similar issue when I run python setup.py build_ext --inplace in dcn. Could you explain in depth how to fix this issue

Hi, if the vesion of pytorch higher than 1.3, you need replace "AT_CHECK" with "TORCH_CHECK" in DB/assets/ops/dcn/src/deform_conv_cuda.cpp and DB/assets/ops/dcn/src/deform_pool_cuda.cpp. After that, you can successfully run “python setup.py build_ext --inplace”