Error: attempt to index field 'CudaByteStorage

[marek094]

Hello, I am trying to run gvnn on AWS Deep Learning Ubuntu, but receive this error in torch:

th> require 'gvnn'
/home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:389: /home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:389: /home/ubuntu/torch/install/share/lua/5.2/cutorch/init.lua:4: attempt to index field 'CudaByteStorage' (a nil value)
stack traceback:
    /home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:506: in function </home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:499>
    [C]: in function 'error'
    /home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:389: in function 'require'
    [string "_RESULT={require 'gvnn'}"]:1: in main chunk
    [C]: in function 'xpcall'
    /home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:661: in function 'repl'
    ...untu/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:204: in main chunk
    [C]: in ?   
                                                                      [0.0008s] 

[ankurhanda]

you need to have cutorch and cudnn and cunn all installed.

[marek094]

Thanks for your support, but even with all those rocks installed the problem occurs. I dont want you to debug my installation but it seems like that I did all right but does not work. I use lua5.2 instead of luaJIT could it be a problem?

ubuntu@ip-172-31-37-80:~/torch/install/bin$ ./luarocks list | egrep '^[a-z]'
Warning: Failed loading manifest for /home/ubuntu/.luarocks/lib/luarocks/rocks: /home/ubuntu/.luarocks/lib/luarocks/rocks/manifest: No such file or directory
argcheck
cudnn
cunn
cutorch
cwrap
dok
env
gnuplot
graph
gvnn
image
lua-cjson
luaffi
luafilesystem
moses
nn
nngraph
nnx
optim
paths
penlight
qtlua
qttorch
sundown
sys
threads
torch
trepl
xlua
ubuntu@ip-172-31-37-80:~/torch/install/bin$ ./th 

  ______             __   |  Torch7 
 /_  __/__  ________/ /   |  Scientific computing for Lua. 
  / / / _ \/ __/ __/ _ \  |  Type ? for help 
 /_/  \___/_/  \__/_//_/  |  https://github.com/torch 
                          |  http://torch.ch 

th> require 'gvnn'
THCudaCheck FAIL file=/tmp/luarocks_cutorch-scm-1-9941/cutorch/lib/THC/THCGeneral.c line=70 error=30 : unknown error
/home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:389: /home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:389: cuda runtime error (30) : unknown error at /tmp/luarocks_cutorch-scm-1-9941/cutorch/lib/THC/THCGeneral.c:70
stack traceback:
    /home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:506: in function </home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:499>
    [C]: in function 'error'
    /home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:389: in function 'require'
    [string "_RESULT={require 'gvnn'}"]:1: in main chunk
    [C]: in function 'xpcall'
    /home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:661: in function 'repl'
    ...untu/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:204: in main chunk
    [C]: in ?   
                                                                      [0.1018s] 
th> require 'cunn'
/home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:389: /home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:389: /home/ubuntu/torch/install/share/lua/5.2/cutorch/init.lua:4: attempt to index field 'CudaByteStorage' (a nil value)
stack traceback:
    /home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:506: in function </home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:499>
    [C]: in function 'error'
    /home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:389: in function 'require'
    [string "_RESULT={require 'cunn'}"]:1: in main chunk
    [C]: in function 'xpcall'
    /home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:661: in function 'repl'
    ...untu/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:204: in main chunk
    [C]: in ?   
                                                                      [0.0011s] 
th> require 'cudnn'
/home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:389: /home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:389: /home/ubuntu/torch/install/share/lua/5.2/cutorch/init.lua:4: attempt to index field 'CudaByteStorage' (a nil value)
stack traceback:
    /home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:506: in function </home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:499>
    [C]: in function 'error'
    /home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:389: in function 'require'
    [string "_RESULT={require 'cudnn'}"]:1: in main chunk
    [C]: in function 'xpcall'
    /home/ubuntu/torch/install/share/lua/5.2/trepl/init.lua:661: in function 'repl'
    ...untu/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:204: in main chunk
    [C]: in ?   
                                                                      [0.0026s] 
th> require 'nn'
{
  SpatialZeroPadding : {...}
  Convert : {...}
  L1Cost : {...}
  Parallel : {...}
  VolumetricFullConvolution : {...}

....

  LookupTable : {...}
  ReLU : {...}
  Threshold : {...}
  Log : {...}
  CReLU : {...}
}
                                                                      [0.0020s] 

[ankurhanda]

It may be that you have to install older version of cudnn. This code was written in June 2016 or around that time. I think a lot of things have changed since then. What do you want to do with it? I can then recommend few things.

[marek094]

Ok, now I will try to build all from scratch again and than with older cudnn.

I am fascinated by your article [1] about Self-supervised learning. It looks like a great bootstrap for many real world problems. I am focusing in my thesis on generating disparity maps from scenes of tens copies of one object. So I would like to repeat your experiment and try to improve it of this domain-specific scenes.

1. https://arxiv.org/abs/1705.08260

EDIT: this worked for me https://github.com/soumith/cudnn.torch/issues/383