run the executable on the PC without Deep Learning Toolbox installed

I installed Matlab R2022a on my PC, Deep Learning Toolbox is also installed. I used the functions of Deep Learning Toolbox in my code, then build a Matlab executable.

I want to run this Matlab executable on another PC, Matlab Runtime R2022a is installed on this PC, I found Deep Learning Toolbox is not installed, Matlab executable crash when running on this PC.

Do you know how to install Deep Learning Toolbox in Matlab Runtime? or how can I run the executable on the PC without Deep Learning Toolbox installed?

Answer 1

[Walter Roberson]

You cannot compile any of the learning / training functions of Deep Learning Toolbox.

You can however train a network beforehand, save the network to a .mat file, and then construct different code that load()'s the network and uses it for prediction or classification, and compile that code.

[bo]

Thanks your help.

I save the imported network to a MAT file.

nnet = importONNXLayers(modelPath, ...);
    nnetwork = assembleNetwork(nnet);
save("model.mat","nnetwork");

Then use .mat for prediction:

load("model.mat","nnetwork");
out = predict(nnetwork,data);

While predict is a function of Deep Learning Toolbox.

I want to do prediction on the PC without Deep Learning Toolbox installed.

[Walter Roberson]

When you use that form of load(), the variable "poofs" into existence, and MATLAB does not know it has been created. You should use

%#function dlnetwork
datacell = load("model.mat","nnetwork");
nnetwork = datacell.nnetwork;

It is possible that you might need to name some other class on the %#function line.

Note that if you want to use predict() then you need to do

nnet = importONNXLayers(modelPath, ...);
nnetwork = dlnetwork(nnet);
save("model.mat","nnetwork");

Using assembleNetwork like you did would be for the case where you planned to use classify()

[bo]

thanks your help.

Is there a way to install the deep learning toolbox in Runtime?

[Steven Lord]

See your other question about the MATLAB Runtime.