How to save training plot in batch mode (deep learning toolbox)

Hi,

In regards to the deep learning toolbox, is there a way to save the training plots to a .mat, .jpg, or .fig?

My issue is that I am running my matlab scripts on a server in batch mode and won't actually see this plot (everything is over the terminal, run with batch files). However I can save data to a .mat file so that I can view it later. Is it possible to save the plot to as a file? Or, maybe there is a better way to do this?

Similar question, is there a way to re-generate the training progress plot after I have finished training the network if I have the net or DAGNetwork file?

Thank you so much for helping me out with this :)

Answer 1

[James Latshaw]

I found an answer here: https://www.mathworks.com/matlabcentral/answers/373963-how-do-i-save-the-training-progress-images-generated-in-trainnetwork-when-plots-is-set-to-train

However, this process can cause output errors if you have very large networks (using lots of memory) because you are trying to save a rich in data plot to memory. I found a work around as plotting the last value of training. So, this isn't a nice plot, but at least it give me some metric to evaluate the RMSE value (or accuracy):

function stop=savetrainingplot(info)

stop=false; %prevents this function from ending trainNetwork prematurely

if info.State=='done' %check if all iterations have completed

% if true

%saveas(gcf,'filename.png') % save figure as .png, you can change this

file_name = ['big3_performance_' num2str(floor(now*100000)) '.mat'];

% to_save_fig = findall(groot, 'Type', 'Figure');

% to_save_fig_last = to_save_fig(end);

%saveas(figure(to_save_fig(end)), file_name)

performance = [info.Iteration info.Epoch info.ValidationRMSE];

save(file_name, 'performance', '-v7.3')

end

end

I hope that this helps someone :)

[Théo]

Hello,

After struggling for a few hours, I found a workaround :

savefig(currentfig, yourfig.fig, 'compact')

The display is a bit affected but still fine.

I found this parameter scrolling down the savefig help document.

Best,

Théo