Unable to save session in Classification Learner

I have trained models in classification learner based on a dataset and am hoping to save the classificationLearner Session so that I do not have to rerun all the models again everytime, which takes quite a long time. I am currently using the R2022a version, and when I try to save the error, I get an error message that says "The file could not be closed, and might now be corrupt." Any advice on how I can get rid of this error and save the session?

[Walter Roberson]

Is it possible that you ran out of disk space?

[Sonia Tan]

The device that I tried it on had >100GB of space so I think there was still sufficient disk space for storage?

Answer 1

[Image Analyst]

What do you mean by "rerun the models"? Why do you have to do that? Did your training data change? If not, why rerun?

What I do is to load by predictor table and ground truth responses and train the model. Then I click the Export->Compact Model to put the "trainedModel" variable into the workspace. Then I call save

save('myModel.mat', 'trainedModel');

to save the trained model to disk. Then to apply the model to predict (not train) new estimated response values, I call load

s = load('myModel.mat');
trainedModel = s.trainedModel;

then call predict() with a table of new data and the trainedModel variable. That will give me an estimated value for each new test value.

[Sonia Tan]

Hi, I was hoping to save the entire classification learner session as my work involves comparing the performance between different models.

I could save the models by exporting each model, but that would mean I have to save each model individually, and I was hoping to save the entire classification learner session so that I do not need to individually save every model.

[Image Analyst]

So what happens when you click the Save button on the tool ribbon of Classification Learner? Can you save it? And then open it again with the Open button?

[Sonia Tan]

nope, I am not able to save it. When I click save, I am prompted to choose a location to save it to, and after choosing the location I will be faced with the error of "The file could not be closed, and might now be corrupt.".

[Image Analyst]

I'd call tech support then.

[Sonia Tan]

okay, thank you!

Answer 2

[Antonio]

In my case, this problem is related to MATLAB having problems saving files larger than 2GB in the default configuration of the "save" function. You can try two options:

• Export the model as a Compact Model (Export -> Export Compact Model) to the workspace and save from there as normally done with save(). If that compact model meets the 2GB limit, you will have no problem.
• Otherwise, export the complete or compact model from the workspace using the option '-v7.3' inside the save() function. It was this option that solved the problem in my case.

save(filename, 'modelTrained', '-v7.3')

I hope to be helpful! Sorry I'm late, but I respond for others who have the same problem.

Information about the '-v7.3' configuration: https://es.mathworks.com/matlabcentral/answers/15521-matlab-function-save-and-v7-3