Error in vl_compilenn (line 506)

Hi guys,

I am trying to use pre-trained convolutional neural network which needs to compile the CUDA for enabling the GPU. I have installed CUDA v8.0 with MATLAB2017a, and downloaded the CUDNN of the same specs, respectively. Everytime I run the vl_compilenn I get the error "Operands to the and && operators must be convertible to logical scalar values. Error in vl_compilenn line(506) if strcmp(arch,'win64') && opts.enableCudnn". Does anybody facing the same problem or have any body solved this issue?

[Joss Knight]

This is a question about MatConvNet - they have their own forums, for instance this Google group, so you may want to start there.

[Junaid Lodhi]

Have you copied the CUDNN library files to the MATLAB default directory/location?

What is the architecture(win32/win64) of MATLAB version 2017a you installed? And, Try to match the architect of your windows OS and MATLAB version.

Answer 1

[WideAwaken]

Please check whether you use the right format of name/value pairs for cudnn. The right name/value pair for cudnn is "...,'enableCudnn', true, ...", rather than "...,'enableCudnn','true',..." . Use the logical value true (no single quotes) , not the string 'true' .