chanyn/Reasoning-RCNN

gcc failed with exit status 1

Opened this issue · 5 comments

Glenn-zkl commented

when i run 'compils.sh' ,it goes with gcc failed with exit staus 1.
anyone come over this problem?

Fran0K commented

hi Glenn,I met this problem few days ago .here is my solutions and hope it can help you out

My error information :
cc1plus: warning: command line option '-Wstrict-prototypes' is valid for C/ObjC but not for C++ src/deform_conv_cuda.cpp: In function 'void shape_check(at::Tensor, at::Tensor, at::Tensor*, at::Tensor, int, int, int, int, int, int, int, int, int, int)': src/deform_conv_cuda.cpp:71:33: error: 'AT_CHECK' was not declared in this scope weight.ndimension());
It keeps repeating the AT_CHECKerror and raises error: command 'gcc' failed with exit status 1at the end

Why:
AT_CHECK is deprecated in pytorch version >=0.4(I know the code should be ran in pyotch 0.4 in author recommendation ,but i ran it in pytorch version =0.4 in docker ,it raise the same gcc error info(sad face here). then I decide to run the code in the pytorch version = 1.9)

My solution:
Replaced AT_CHECK with TORCH_CHECK in the following files
Reasoning-RCNN/mmdet/ops/roi_align/src/roi_align_cuda.cpp
Reasoning-RCNN/mmdet/ops/roi_pool/src/roi_pool_cuda.cpp
Reasoning-RCNN/mmdet/ops/dcn/src/deform_conv_cuda.cpp
Reasoning-RCNN/mmdet/ops/dcn/src/deform_pool_cuda.cpp

run./compile.sh and 'gcc' failed with exit status 1 disappear at the end（btw My way work fine in pytorch = 1.9 in docker )

QingtianK commented

Hi Fran0K and Glenn .Did you two run this code successfully?

Glenn-zkl commented

Hi Fran0K and Glenn .Did you two run this code successfully?

Hi Fran0K and Glenn .Did you two run this code successfully?

not yet,because i was trying to realize other method on my own baseline before now.
and now i am suffering from another problem that is how to generate a 'graph.pkl' used on my own dataset.
So,sir,do you know how to generate it?
: )

Glenn-zkl commented

hi Glenn,I met this problem few days ago .here is my solutions and hope it can help you out

My error information :
cc1plus: warning: command line option '-Wstrict-prototypes' is valid for C/ObjC but not for C++ src/deform_conv_cuda.cpp: In function 'void shape_check(at::Tensor, at::Tensor, at::Tensor*, at::Tensor, int, int, int, int, int, int, int, int, int, int)': src/deform_conv_cuda.cpp:71:33: error: 'AT_CHECK' was not declared in this scope weight.ndimension());
It keeps repeating the AT_CHECKerror and raises error: command 'gcc' failed with exit status 1at the end

Why:
AT_CHECK is deprecated in pytorch version >=0.4(I know the code should be ran in pyotch 0.4 in author recommendation ,but i ran it in pytorch version =0.4 in docker ,it raise the same gcc error info(sad face here). then I decide to run the code in the pytorch version = 1.9)

My solution:
Replaced AT_CHECK with TORCH_CHECK in the following files
Reasoning-RCNN/mmdet/ops/roi_align/src/roi_align_cuda.cpp
Reasoning-RCNN/mmdet/ops/roi_pool/src/roi_pool_cuda.cpp
Reasoning-RCNN/mmdet/ops/dcn/src/deform_conv_cuda.cpp
Reasoning-RCNN/mmdet/ops/dcn/src/deform_pool_cuda.cpp

run./compile.sh and 'gcc' failed with exit status 1 disappear at the end（btw My way work fine in pytorch = 1.9 in docker )

Thanks for your suggestion. ： ）
According to what you say , i guess you run it successfully. So, do you know how to generate the 'graph.pkl' ? This is not interpreted in the paper.

QingtianK commented

Hi Fran0K and Glenn .Did you two run this code successfully?

Hi Fran0K and Glenn .Did you two run this code successfully?

not yet,because i was trying to realize other method on my own baseline before now.
and now i am suffering from another problem that is how to generate a 'graph.pkl' used on my own dataset.
So,sir,do you know how to generate it?
: )

Sorry，I have the same problem.