Open guiji0812 opened 5 years ago
Hi,Thank you very much for sharing! My pytorch version is 1.0.1,cuda=9.0. I can install the relevant package for this code. I can also successfully run "bash make.sh". However, running "python test_Vid4_REDS4_with_GT.py" has the following error: "ImportError: Failed to import DCNv2 module". I made the relevant changes, but it didn't work. What is the correct way to open?
If with this error "ImportError: Failed to import DCNv2 module
, it means the installation of DCN failed.
Hi @xinntao,
I did invoke make.sh command in /models/modules/DCNv2. The result is
Installed ./codes/models/modules/DCNv2 Processing dependencies for DCNv2==0.1 Finished processing dependencies for DCNv2==0.1
However, I still encountered this problem,
ImportError: Failed to import DCNv2 module
What could be the problem? Thank you.
Hi @xinntao,
I did invoke make.sh command in /models/modules/DCNv2. The result is
Installed ./codes/models/modules/DCNv2 Processing dependencies for DCNv2==0.1 Finished processing dependencies for DCNv2==0.1
However, I still encountered this problem,
ImportError: Failed to import DCNv2 module
What could be the problem? Thank you.
I have used several hours to solve this problem. It caused by your gcc compiler version. If your gcc compiler version lower than 4.9, please update to 4.9 or higher. I recommend you use 5.4.0, because I update my gcc compiler to 7.3.0 and I get another "undefined symbol..." error. I didn't want to reinstall my gcc compiler because it will cost 1~2hours. So, I solved the new error by this: https://github.com/SeanNaren/warp-ctc/issues/103#issuecomment-507138951
Hi @xinntao, I did invoke make.sh command in /models/modules/DCNv2. The result is
Installed ./codes/models/modules/DCNv2 Processing dependencies for DCNv2==0.1 Finished processing dependencies for DCNv2==0.1
However, I still encountered this problem,
ImportError: Failed to import DCNv2 module
What could be the problem? Thank you.
I have used several hours to solve this problem. It caused by your gcc compiler version. If your gcc compiler version lower than 4.9, please update to 4.9 or higher. I recommend you use 5.4.0, because I update my gcc compiler to 7.3.0 and I get another "undefined symbol..." error. I didn't want to reinstall my gcc compiler because it will cost 1~2hours. So, I solved the new error by this: SeanNaren/warp-ctc#103 (comment)
Thank you, @siyux. However, I already have the gcc-5.4.0 version. The test_Vid4_REDS4_with_GT_DUF.py is working perfectly but the test_Vid4_REDS4_with_GT.py is encountering the above stated problem.
Hi @xinntao, I did invoke make.sh command in /models/modules/DCNv2. The result is
Installed ./codes/models/modules/DCNv2 Processing dependencies for DCNv2==0.1 Finished processing dependencies for DCNv2==0.1
However, I still encountered this problem,
ImportError: Failed to import DCNv2 module
What could be the problem? Thank you.
I have used several hours to solve this problem. It caused by your gcc compiler version. If your gcc compiler version lower than 4.9, please update to 4.9 or higher. I recommend you use 5.4.0, because I update my gcc compiler to 7.3.0 and I get another "undefined symbol..." error. I didn't want to reinstall my gcc compiler because it will cost 1~2hours. So, I solved the new error by this: SeanNaren/warp-ctc#103 (comment)
Thank you, @siyux. However, I already have the gcc-5.4.0 version. The test_Vid4_REDS4_with_GT_DUF.py is working perfectly but the test_Vid4_REDS4_with_GT.py is encountering the above stated problem.
I have a similar problem. So, I recompiled DCNv2 and everything goes well! Hope it also can help you.
Hi @xinntao, I did invoke make.sh command in /models/modules/DCNv2. The result is
Installed ./codes/models/modules/DCNv2 Processing dependencies for DCNv2==0.1 Finished processing dependencies for DCNv2==0.1
However, I still encountered this problem,
ImportError: Failed to import DCNv2 module
What could be the problem? Thank you.
I have used several hours to solve this problem. It caused by your gcc compiler version. If your gcc compiler version lower than 4.9, please update to 4.9 or higher. I recommend you use 5.4.0, because I update my gcc compiler to 7.3.0 and I get another "undefined symbol..." error. I didn't want to reinstall my gcc compiler because it will cost 1~2hours. So, I solved the new error by this: SeanNaren/warp-ctc#103 (comment)
Thank you, @siyux. However, I already have the gcc-5.4.0 version. The test_Vid4_REDS4_with_GT_DUF.py is working perfectly but the test_Vid4_REDS4_with_GT.py is encountering the above stated problem.
I have a similar problem. So, I recompiled DCNv2 and everything goes well! Hope it also can help you.
You mean delete "build" folder and re-run python setup.py develop? I can not fix this problem in this way. It seems it doesn't work for torch1.3 and cuda10. It works when I use torch1.1 and cuda9.0
hi, when I test my own images, the image size can't be divided by 4,then the size in PCD Align Module will be different between refference image and the neighbor image,how to solve this problem?