2_unsupervised and CUDA

Question

2_unsupervised and CUDA

markpeot opened this issue 11 years ago · comments

if I run
torch -i doall.lua -type cuda

I get this error:
torch-qlua: /usr/local/share/torch/lua/nn/SpatialConvolutionMM.lua:42: attempt to call field 'SpatialConvolutionMM_updateOutput' (a nil value)
stack traceback:
[C]: in function 'SpatialConvolutionMM_updateOutput'
/usr/local/share/torch/lua/nn/SpatialConvolutionMM.lua:42: in function 'updateOutput'
/usr/local/share/torch/lua/nn/Sequential.lua:27: in function 'forward'
4_train.lua:157: in function 'opfunc'
/usr/local/share/torch/lua/optim/sgd.lua:36: in function 'optimMethod'
4_train.lua:181: in function 'train'
doall.lua:76: in main chunk

The code works fine if I use the default arguments.
Should I be using a different version of the tutorial?

Clement Farabet · Answer 1 · Tue Oct 29 2013 21:04:14 GMT+0800 (China Standard Time)

The tutorials are not really compatible with the GPU code anymore. I'll have to update this code.

Sergey Zagoruyko · Answer 2 · Thu Jul 31 2014 07:57:03 GMT+0800 (China Standard Time)

@clementfarabet I've got the same error with supervised tutorial.

Clement Farabet · Answer 3 · Thu Jul 31 2014 12:06:30 GMT+0800 (China Standard Time)

Can you guys try to re-install cutorch and cunn?

Sergey Zagoruyko · Answer 4 · Thu Jul 31 2014 17:52:24 GMT+0800 (China Standard Time)

@clementfarabet I did luarocks remove cutorch and luarocks remove cunn and installed it again, got the same error. Tried on 2 machines, mac and ubuntu.

Soumith Chintala · Answer 5 · Thu Oct 09 2014 12:33:10 GMT+0800 (China Standard Time)

fixed 2_supervised to be CUDA compatible, and just tested it.