编译安装遇到的问题和解决方法记录

[XksA-me]

环境：ubuntu18.04 cuda10.0 py3.6 torch1.4.0

1、ModuleNotFoundError: No module named 'Cython'

pip install Cython

2、/PyTorch-Encoding/encoding/lib/gpu/lib_ssd.cu(144): error: identifier "FLT_MAX" is undefined
/PyTorch-Encoding/encoding/lib/gpu/lib_ssd.cu 代码开头加：

#include <cfloat>

3、/PyTorch-Encoding/encoding/lib/gpu/common.h(10): error: class template "ScalarConvert" has already been defined
/PyTorch-Encoding/encoding/lib/gpu/common.h 这个文件最开头 加下面代码

没解决

[XiShuFan]

Have you solved the 3rd problem?
encoding/lib/gpu/common.h(10): error: class template "ScalarConvert" has already been defined

I'm looking forward to your answer!

[TiSgrc]

Have you solved the 3rd problem? encoding/lib/gpu/common.h(10): error: class template "ScalarConvert" has already been defined

I'm looking forward to your answer!

I had this problem too, have you solved it?

[crankler]

Same issue, have you solve it now?

[CXavireH]

Have you solved the 3rd problem? encoding/lib/gpu/common.h(10): error: class template "ScalarConvert" has already been defined
I'm looking forward to your answer!

I had this problem too, have you solved it?

same problem, have you solve it?

[XIANGLONGGUO]

Have you solved the 3rd problem? encoding/lib/gpu/common.h(10): error: class template "ScalarConvert" has already been defined
I'm looking forward to your answer!

I had this problem too, have you solved it?

same problem, have you solve it?

I solved the 3rd problem by remove '#include "common.h"" in 'encoding/lib/gpu/rectify_cuda.cu'.