Memory issue

[vidda13]

I am trying to run the network on GTX 720 and i get the following issue
torch/install/share/lua/5.1/nn/utils.lua:11: cuda runtime error (2) : out of memory at /tmp/luarocks_cutorch-scm-1-2463/cutorch/lib/THC/THCStorage.cu:44
stack traceback:
[C]: in function 'resize'
torch/install/share/lua/5.1/nn/utils.lua:11: in function 'torch_Storage_type'
torch/install/share/lua/5.1/nn/utils.lua:57: in function 'recursiveType'
torch/install/share/lua/5.1/nn/Module.lua:123: in function 'type

I tried using smaller image_size and went to a size of 28 too. But it still doesn't work. Any suggestions what parameters I can tweak?

[hughperkins]

I have a 1GB card, and runs just fine. I'm using the normalised vgg 19 though, not the original version. normalised hsa no fc layers, which saves 450MB or so. (ps, I can train up to size 200 just fine, with 1GB; size 256 runs out of memory for me; using lbfgs)

[vidda13]

ok.. did you use the same hyper-parameters? The GTX 720 has 2GB. Any idea how much memory reducing an image size can save?

[hughperkins]

try these: #44 (comment)

th neural_style.lua -style_image examples/inputs/picasso_selfport1907.jpg -content_image examples/inputs/brad_pitt.jpg -gpu 0 -output_image profile.png -image_size 200 -model_file models/vgg_normalised.caffemodel -num_iterations 1000 -save_iter 50 -normalize_gradients -content_weight 50000 -style_weight 90000

[vidda13]

Well I guess the issue lies elsewhere. As i get the same error at cuda runtime error (2) : out of memory at /tmp/luarocks_cutorch-scm-1-2463/cutorch/lib/THC/THCStorage.cu:44.

[hughperkins]

Ah, then I'm out of ideas. (other than basic things like: did you make sure to upgrade all of: nn, torch7, cutorch, cunn, inn? are you using cuda7?) (oh wait, is this gpu being used for your display/monitor too? is it possible that your monitor usage is somehow using all your memory? maybe you can use nvidia-smi to check how much memory is available when neural_style is not runnign?)

[vidda13]

Yikes. Some process was clogging 1.5GB of the memory. Seems to be running fine now when I restarted my system. Thanks.

[hughperkins]

:-)