Neural-style: Error when running without CUDA: No LuaRocks module found for cutorch

[cpury]

Trying to run with the default parameters. Getting this error:

/Users/x/torch/install/bin/luajit: /Users/x/torch/install/share/lua/5.1/trepl/init.lua:363: module 'cutorch' not found:No LuaRocks module found for cutorch
    no field package.preload['cutorch']
    no file '/Users/x/.luarocks/share/lua/5.1/cutorch.lua'
    no file '/Users/x/.luarocks/share/lua/5.1/cutorch/init.lua'
    no file '/Users/x/torch/install/share/lua/5.1/cutorch.lua'
    no file '/Users/x/torch/install/share/lua/5.1/cutorch/init.lua'
    no file './cutorch.lua'
    no file '/Users/x/torch/install/share/luajit-2.1.0-alpha/cutorch.lua'
    no file '/usr/local/share/lua/5.1/cutorch.lua'
    no file '/usr/local/share/lua/5.1/cutorch/init.lua'
    no file '/Users/x/.luarocks/lib/lua/5.1/cutorch.so'
    no file '/Users/x/torch/install/lib/lua/5.1/cutorch.so'
    no file './cutorch.so'
    no file '/usr/local/lib/lua/5.1/cutorch.so'
    no file '/usr/local/lib/lua/5.1/loadall.so'
stack traceback:
    [C]: in function 'error'
    /Users/x/torch/install/share/lua/5.1/trepl/init.lua:363: in function 'require'
    neural_style.lua:42: in function 'main'
    neural_style.lua:370: in main chunk
    [C]: in function 'dofile'
    .../x/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:131: in main chunk
    [C]: at 0x01028fc7b0

I tried explicitly using the nn engine. Also to install cutorch, which fails due to Specify CUDA_TOOLKIT_ROOT_DIR.

Note, this is OSX.

Thanks!

[jcjohnson]

If you don't have cutorch installed then you won't be able to run in GPU mode. You can run in CPU only mode by passing the flag -gpu -1.

[cpury]

Ah missed that line!

Ok now I'm getting

[...]
Successfully loaded models/VGG_ILSVRC_19_layers.caffemodel
/Users/x/torch/install/bin/luajit: models/VGG_ILSVRC_19_layers_deploy.prototxt.cpu.lua:7: attempt to call method 'ceil' (a nil value)
stack traceback:
    models/VGG_ILSVRC_19_layers_deploy.prototxt.cpu.lua:7: in main chunk
    [C]: in function 'dofile'
    ./loadcaffe_wrapper.lua:48: in function 'load'
    neural_style.lua:53: in function 'main'
    neural_style.lua:370: in main chunk
    [C]: in function 'dofile'
    .../x/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:131: in main chunk
    [C]: at 0x010919c7b0

[jcjohnson]

That should be fixed if you update your nn to the latest version; the ceil() method was added pretty recently in https://github.com/torch/nn/commit/929cfc57c88952b597bec77046582b90d1122380.

[cpury]

Ha, thanks man that worked!! :)

[mintisan]

@jcjohnson
how do i install cutorck, I execute luarocks install cutorch,but i encounter error below

[jcjohnson]

Do you have CUDA installed?

[mintisan]

@jcjohnson
like brew cask install cuda?

OS : Mac

[jcjohnson]

I don't think you can install cuda with homebrew; try following the steps here:

http://docs.nvidia.com/cuda/cuda-getting-started-guide-for-mac-os-x/#axzz3km0bJe3r

Alternatively you can run neural-style in CPU only mode with the flag -gpu -1.

[mintisan]

@jcjohnson Thanks , I will have a try.

ps : in CPU mode is too slow :D

[reillydonovan]

I'm receiving the same error, however CUDA is installed. I can get CUDA to work with DeepDreamVideo and when I type:

nvidia-smi
Thu Oct 1 17:17:11 2015
+------------------------------------------------------+
| NVIDIA-SMI 346.46 Driver Version: 346.46 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
|===============================+======================+======================|
| 0 GRID K520 Off | 0000:00:03.0 Off | N/A |
| N/A 32C P0 37W / 125W | 10MiB / 4095MiB | 0% Default |
+-------------------------------+----------------------+----------------------+

+-----------------------------------------------------------------------------+
| Processes: GPU Memory |
| GPU PID Type Process name Usage |
|=============================================================================|
| No running processes found |
+-----------------------------------------------------------------------------+

However, when I try to install cutorch, I'm returned:
luarocks install cutorch
Installing https://raw.githubusercontent.com/torch/rocks/master/cutorch-scm-1.rockspec...
Using https://raw.githubusercontent.com/torch/rocks/master/cutorch-scm-1.rockspec... switching to 'build' mode
Cloning into 'cutorch'...
remote: Counting objects: 81, done.
remote: Compressing objects: 100% (77/77), done.
remote: Total 81 (delta 7), reused 25 (delta 0), pack-reused 0
Receiving objects: 100% (81/81), 122.26 KiB | 0 bytes/s, done.
Resolving deltas: 100% (7/7), done.
Checking connectivity... done.
cmake -E make_directory build && cd build && cmake .. -DCMAKE_BUILD_TYPE=Release -DCMAKE_PREFIX_PATH="/home/ubuntu/torch/install/bin/.." -DCMAKE_INSTALL_PREFIX="/home/ubuntu/torch/install/lib/luarocks/rocks/cutorch/scm-1" && make -j$(getconf _NPROCESSORS_ONLN) install

-- The C compiler identification is GNU 4.8.4
-- The CXX compiler identification is GNU 4.8.4
-- Check for working C compiler: /usr/bin/cc
-- Check for working C compiler: /usr/bin/cc -- works
-- Detecting C compiler ABI info
-- Detecting C compiler ABI info - done
-- Check for working CXX compiler: /usr/bin/c++
-- Check for working CXX compiler: /usr/bin/c++ -- works
-- Detecting CXX compiler ABI info
-- Detecting CXX compiler ABI info - done
-- Found Torch7 in /home/ubuntu/torch/install
CMake Error at /usr/share/cmake-2.8/Modules/FindPackageHandleStandardArgs.cmake:108 (message):
Could NOT find CUDA (missing: CUDA_INCLUDE_DIRS CUDA_CUDART_LIBRARY) (found
suitable version "7.0", minimum required is "5.5")
Call Stack (most recent call first):
/usr/share/cmake-2.8/Modules/FindPackageHandleStandardArgs.cmake:315 (_FPHSA_FAILURE_MESSAGE)
/usr/share/cmake-2.8/Modules/FindCUDA.cmake:806 (find_package_handle_standard_args)
CMakeLists.txt:7 (FIND_PACKAGE)

-- Configuring incomplete, errors occurred!
See also "/tmp/luarocks_cutorch-scm-1-6030/cutorch/build/CMakeFiles/CMakeOutput.log".

Error: Build error: Failed building.

Any suggestions on how to properly install cutorch? It seems like it can find CUDA but just isn't using it because of some kind of directory issue

[kirilledelman]

Did you do this step? ( your actual dir will vary )

Set up the required environment variables:
export PATH=/Developer/NVIDIA/CUDA-7.0/bin:$PATH
export DYLD_LIBRARY_PATH=/Developer/NVIDIA/CUDA-7.0/lib:$DYLD_LIBRARY_PATH

[lihuibng]

run the command below of densecap:
th run_model.lua -input_image imgs/elephant.jpg -gpu -1
got error :
/Users/xxx/torch/install/bin/lua: /Users/wuyuangui/torch/install/share/lua/5.2/trepl/init.lua:384: /Users/xxx/torch/install/share/lua/5.2/trepl/init.lua:384: /Users/xxx/torch/install/share/lua/5.2/trepl/init.lua:384: module 'cutorch' not found:No LuaRocks module found for cutorch
no field package.preload['cutorch']
no file '/Users/xxx/.luarocks/share/lua/5.2/cutorch.lua'
no file '/Users/xxx/.luarocks/share/lua/5.2/cutorch/init.lua'
no file '/Users/xxx/torch/install/share/lua/5.2/cutorch.lua'
no file '/Users/xxx/torch/install/share/lua/5.2/cutorch/init.lua'
no file '/Users/xxx/.luarocks/share/lua/5.1/cutorch.lua'
no file '/Users/xxx/.luarocks/share/lua/5.1/cutorch/init.lua'
no file '/Users/xxx/torch/install/share/lua/5.1/cutorch.lua'
no file '/Users/xxx/torch/install/share/lua/5.1/cutorch/init.lua'
no file './cutorch.lua'
no file '/Users/xxx/torch/install/share/luajit-2.1.0-beta1/cutorch.lua'
no file '/usr/local/share/lua/5.1/cutorch.lua'
no file '/usr/local/share/lua/5.1/cutorch/init.lua'
no file '/Users/xxx/.luarocks/lib/lua/5.2/cutorch.so'
no file '/Users/xxx/torch/install/lib/lua/5.2/cutorch.so'
no file '/Users/xxx/torch/install/lib/cutorch.dylib'
no file '/Users/xxx/.luarocks/lib/lua/5.1/cutorch.so'
no file '/Users/xxx/torch/install/lib/lua/5.1/cutorch.so'
no file './cutorch.so'
no file '/usr/local/lib/lua/5.1/cutorch.so'
no file '/usr/local/lib/lua/5.1/loadall.so'
stack traceback:
[C]: in function 'error'
/Users/xxx/torch/install/share/lua/5.2/trepl/init.lua:384: in function 'require'
run_model.lua:5: in main chunk
[C]: in function 'dofile'
...ngui/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:145: in main chunk
[C]: in ?

any idea ?
my Mac got the Inter GPu, so I want to run the command in CPU only models.

[jmbeach]

Yeah I would appreciate mac help.

I'm downloading cuda now from here.

I'm thinking if it still doesn't work, there's nothing stopping me from editing this file: /usr/local/Cellar/cmake/3.3.1/share/cmake/Modules/FindCUDA.cmake and manually setting the variable:

set(CUDA_TOOLKIT_ROOT_DIR "/usr/local/cuda/")

or something like that

[jmbeach]

Also, following these directions: http://docs.nvidia.com/cuda/cuda-getting-started-guide-for-mac-os-x/#axzz474n2mSdf

[jmbeach]

Ok, everyone just download from the above installer and follow those directions. That did the trick for me (at least got me past that error)

[lihuibng]

but I did not got a NVIDIA GPU, the GPU is Intel Iris Graphics 6100 1536 MB , I install thie CUDA toolkit and driver , there is another error ' CUDA driver version is insufficient for CUDA runtime version.', any idea?

[jcjohnson]

@lihuibng CUDA won't work on an Intel GPU; it only works on NVIDIA GPUs. OpenCL might work on the Intel GPU, but you probably won't have enough memory for neural-style to work. Your best bet is CPU mode.

[jmbeach]

Yeah, I realized I don't have an Nvidia card either (don't know why I thought I did) and was running in CPU and still got an error about some png library

[vinhqdang]

Hello

Is there a way to automatically select CPU or GPU mode? I have two computers, one with CUDA and one without. I maintained one code repository but I have to modify the code locally.

[sdziscool]

@vinhqdang yes, I wrote a few wiki pages just for people like you

the wiki You can find the instructions under "basic usage" in the "An Introduction to Neural Style" page

[vinhqdang]

Hello @sdziscool

Thanks a lot for pointing me out a very interesting project 👍 However, I think it is not really my question.

I would like to have something like:

if CUDA is supported on the system:
  run with -gpu 0 -gpu_backend cuda (default parameters)
elif OpenCL is supported:
  run with -gpu_backend opencl
else:
  run with -gpu -1

[sdziscool]

@vinhqdang
Mmmh that's not going to be easy, the most basic thing you could do is check if deviceQuery or nvcc and clinfo actually return anything other than an error.

For now you're better of running everything with OpenCL via -backend clnn after you installed all the dependencies. I haven't tested it but it seems like both CPUs and NVIDIA have OpenCL support. You can check if they have OpenCL support by simply running clinfo after having installed the correct drivers.

Such a command would always use the GPU though because 95% of GPUs support OpenCL, so you'd have to check if a server was running headless without GPU or not. basically

elif OpenCL is supported:

will always return true, so you'd need a way to differentiate between an OpenCL enabled CPU and an OpenCL enabled GPU which is another challenge on it's own.

I'm sorry I can't provide you with a concrete answer but I hope i was able to help you somewhat!
Also I'd recommend opening up a new issue instead of interfering with an old one, as your problem is completely unrelated to this issue.

[htoyryla]

I think it should be possible to adapt to which modules have been installed as seen by lua, instead of actual HW capabilities, but have not checked that. Something like discussed here http://stackoverflow.com/questions/15429236/how-to-check-if-a-module-exists-in-lua

But on the other hand a fully automatic selection would be limiting... someone with a GPU with limited RAM may want to run neural-style job with CPU sometimes. Actually I didn't see the problem behind the request. There is no need to "modify code", just to use the correct parameters for -gpu and -backend when running neural-style.

[EliseSo]

@lihuibng have you solved the GPU problem? I got the same error message: CUDA driver version is insufficient for CUDA runtime version. My GPU is Intel Iris too.

[kpranav1998]

th computeProposals.lua $DEEPMASK/pretrained/deepmask -gpu -1
/home/pranav/torch/install/bin/luajit: /home/pranav/torch/install/share/lua/5.1/trepl/init.lua:389: module 'cutorch' not found:No LuaRocks module found for cutorch
no field package.preload['cutorch']
no file '/home/pranav/.luarocks/share/lua/5.1/cutorch.lua'
no file '/home/pranav/.luarocks/share/lua/5.1/cutorch/init.lua'
no file '/home/pranav/torch/install/share/lua/5.1/cutorch.lua'
no file '/home/pranav/torch/install/share/lua/5.1/cutorch/init.lua'
no file './cutorch.lua'
no file '/home/pranav/torch/install/share/luajit-2.1.0-beta1/cutorch.lua'
no file '/usr/local/share/lua/5.1/cutorch.lua'
no file '/usr/local/share/lua/5.1/cutorch/init.lua'
no file '/home/pranav/.luarocks/lib/lua/5.1/cutorch.so'
no file '/home/pranav/torch/install/lib/lua/5.1/cutorch.so'
no file '/home/pranav/torch/install/lib/cutorch.so'
no file './cutorch.so'
no file '/usr/local/lib/lua/5.1/cutorch.so'
no file '/usr/local/lib/lua/5.1/loadall.so'
stack traceback:
[C]: in function 'error'
/home/pranav/torch/install/share/lua/5.1/trepl/init.lua:389: in function 'require'
computeProposals.lua:11: in main chunk
[C]: in function 'dofile'
...anav/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:150: in main chunk
[C]: at 0x00405d50

I am still getting the same error pls help in this case .I am not able to run in cpu mode

[xun6000]

running with flag -gpu -1 still get the same error: no field package.preload['cudnn']