GPU out of Memory on vgg-verydeep-16

[Amir Shehzad]

Hello i am getting GPU out of Memory Error when using vgg-verydeep-16 both as CNN A and CNN B even with batchSize = 64 or 32 or 1.

Can you please tell me how much GPU RAM is required to run this??

My system has 34 GB RAM and I have GPU with 4 GB RAM. I am using Matlab 2017a , cuda 7.5 and ubuntu 14

Error using gpuArray/max
Out of memory on device. To view more detail about available memory on the GPU, use 'gpuDevice()'. If the problem
persists, reset the GPU by calling 'gpuDevice(1)'.

Error in vl_nnrelu (line 38)
    y = max(x, 0) ;

Error in vl_bilinearnn (line 242)
      res(i+1).x = vl_nnrelu(res(i).x,[],leak{:}) ;

Error in initializeNetworkSharedWeights (line 130)
            res = vl_bilinearnn(netInit, im, [], res, ...

Error in imdb_bcnn_train_dag (line 65)
net = initNetFn(imdb, encoderOpts, opts);

Error in run_experiments_bcnn_train (line 79)
          imdb_bcnn_train_dag(imdb, opts);

[tsungyu]

I usually use gpus more than 10gb. Reduce the batchSize at the line 100 in initializeNetworkSharedWeights.m to see if this could work. This part of the code precomputes the bcnn features for initializing the weight for the classification layer.