cjson.so: undefined symbol: lua_objlen stack traceback: train.lua

[Malhabib]

/home/alhabib/torch/install/bin/lua: /home/alhabib/torch/install/share/lua/5.2/trepl/init.lua:389: .../alhabib/torch/install/share/lua/5.2/luarocks/loader.lua:117: error loading module 'cjson' from file '/home/alhabib/torch/install/lib/lua/5.2/cjson.so': /home/alhabib/torch/install/lib/lua/5.2/cjson.so: undefined symbol: lua_objlen stack traceback: [C]: in function 'error' /home/alhabib/torch/install/share/lua/5.2/trepl/init.lua:389: in function 'require' train.lua:14: in main chunk [C]: in function 'dofile' ...abib/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:150: in main chunk [C]: in ? I tried many solutions but non of them works, I'm using Torch 7 and cuda 8.0 , python 2.7, can anyone help me , it happend with (th train.lua) , the preprocessing step works fine.

[aurooj]

I am facing the same problem. Did you figure out the solution?

[aurooj]

This problem was solved by downgrading the cjon version as mentioned here: https://github.com/mpx/lua-cjson/issues/56#issuecomment-394764240

[jainal09]

Hey, Check this docker image I created for running dense cap in gpu! https://hub.docker.com/r/jainal09/densecap_gpu

For Cpu version. Use this docker file. https://github.com/runwayml/alpha_models/blob/master/densecap/Dockerfile