I have generated code for multiple tensorflow-lite networks. I got the .o files. Now I'm trying to build all of them into one c++ application. But I get linker warnings when trying to build the application
/usr/bin/ld: ... XXXX.o):(.bss+0x10): multiple definition ofactivationsBaseAddress'; ... .o):(.bss+0x10): first defined here ...`
/usr/bin/ld: ... YYYY.o):(.bss+0x0): multiple definition ofconstWeightsBaseAddress'; ... .o):(.bss+0x0): first defined here ...`
/usr/bin/ld: ... ZZZZ.o):(.bss+0x8): multiple definition ofmutableWeightsBaseAddress'; ... .o):(.bss+0x8): first defined here ...`
.. and so on.
All .o file seems to export the same symbols. The linker is not happy.
I haven't found any documentation or example of trying to import more than one network. Maybe it's impossible?
I have generated code for multiple tensorflow-lite networks. I got the .o files. Now I'm trying to build all of them into one c++ application. But I get linker warnings when trying to build the application
/usr/bin/ld: ... XXXX.o):(.bss+0x10): multiple definition ofactivationsBaseAddress'; ... .o):(.bss+0x10): first defined here ...`/usr/bin/ld: ... YYYY.o):(.bss+0x0): multiple definition ofconstWeightsBaseAddress'; ... .o):(.bss+0x0): first defined here ...`/usr/bin/ld: ... ZZZZ.o):(.bss+0x8): multiple definition ofmutableWeightsBaseAddress'; ... .o):(.bss+0x8): first defined here ...`.. and so on.
All .o file seems to export the same symbols. The linker is not happy.
I haven't found any documentation or example of trying to import more than one network. Maybe it's impossible?
Is there a way to around this issue?