Issues with cnngpu.dll in NN toolbox

I have been following through the NN toolbox object detection tutorial (https://www.mathworks.com/examples/matlab-computer-vision/mw/vision_product-DeepLearningRCNNObjectDetectionExample-object-detection-using-deep-learning#1), and, while trying to construct my network's layers, I have been running into the following issue:

Error using nnet.internal.cnn.layer.Convolution2D Can't reload 'C:\Program Files\MATLAB\R2016a\bin\win64\cnngpu.dll'

Error in convolution2dLayer (line 76) internalLayer = nnet.internal.cnn.layer.Convolution2D(args.Name, ...

I could really use some help, please!

[Joss Knight]

Try rebooting your machine, and make sure you only have one copy of MATLAB running. Then let me know whether your issue remained.

[Marina Fridin]

I rebooted my machine, and made sure that there is only one open Matlab copy.

I am running Matlab from a virtual machine - could that be a potential issue?

[Joss Knight]

What is the output of

version -modules

And what is the output of

ls(fullfile(matlabroot, 'bin', computer('arch'), '*cnn*'))

[Joss Knight]

Okay, I've looked this up and now I know where this error comes from - see my answer.

[Chow]

Have you solved the issue which you had mentioned?

Answer 1

[Joss Knight]

This is the unfortunate error message you get when you do not have Parallel Computing Toolbox and therefore cannot use the ConvNet feature in R2016a.

[Yicong Xu]

I have the Parallel Computing Toolbox and have cnngpu.dll too, but still cannot use the convolution2dLayer function

[Joss Knight]

Can you confirm this by showing me the output of the command

   license('inuse')

Thanks.

[Chow]

I have the same issue,and I tyied your order,and the output like this.