Why am i getting 'Not enough input arguments?

function out = CodeNNet(in)

persistent NNet
    if isempty(NNet)
        NNet = coder.loadDeepLearningNetwork('NNet.mat');
    end

out = predict(NNet,in);

cfg = coder.config('lib');
cfg.TargetLang = 'C++';
cfg.GenCodeOnly = true;

dlcfg = coder.DeepLearningConfig('arm-compute');
dlcfg.ArmArchitecture = 'armv8';
dlcfg.ArmComputeVersion = '20.02.1';
cfg.DeepLearningConfig = dlcfg;

codegen -args {ones(256,256,3,'single')} -config cfg CodeNNet

[Ignacio Acevedo]

Im trying to generate the C++ code for a trained neural network, but i get an error in line 8 about not enough input arguments

[Jan]

Please post a copy of the complete error message. Which function does not get enough input arguments? predict, or is in a function also, or do you call the function CodeNNet without input arguments, e.g. by starting it using the green triangle in the editor?

[Gaurav Garg]

Hi Igancio,

Can you post a picture of the error you are facing?

[Jan]

I assume, this thread is abandoned.

[John D'Errico]

Poor sad thread, unloved and abandoned.

[Jan]

I do love this thread. I do care and Gaurav does also and you, John, show your emotions as well.

Thread, we are here, if you need us.

[Ignacio Acevedo]

[Ignacio Acevedo]

i understand that i have no input for the function but is it possible to generate C++ code just for the Neural Network itself?

[Walter Roberson]

You need a prototype call to generate code, as code generation specializes the datatypes according to the prototype.

[Ignacio Acevedo]

Like a script where the function is called and tested with a test input?

[Walter Roberson]

The

codegen -args {ones(256,256,3,'single')} -config cfg CodeNNet

should not be part of the source code, and the function name must go before the -args

Answer 1

[Gaurav Garg]

Hi,

From what seems to be the problem so far-

You should not include

codegen -args {ones(256,256,3,'single')} -config cfg CodeNNet

in your source code.