Inconsistent training loss curve when training UNet with and without convergence criterion

I want to train a semantic segmentation model using UNet layers. I trained the model one time without any convergence criterion and max epoch of 500 and the other time with a convergence criterion (validation patience of 10 and validation frequency of 25). The secnd training process converged at a lower epoch number compared to the first one. The results of the application of the second model is significantly less accurate than the first one on the test slices. I noticed that the training loss in the second training process is different from the first one. My question is that why defining a convergence criterion changes the training loss curve? I would be thankful if any one can help me understand why this is happening. The learning rate is constant and equal to 1e-05.

Training curves without convergence criterion:

Training curves with convergence criterion:

Answer 1

[Matt J]

There should be a change in the training loss curve - it should be shorter. By loosening the stopping criteria, you cause fewer iterations to be run.

There will also be changes in the training loss curves due to the stochastic nature of the SGD algorithm and the parameter initialization.

[Memo Remo]

Dear Matt,

Thanks for your reply.

Exactly. It should not change the training loss curve. So the curves above should be the same until the epoch number of 350. However, they are not! The two training processes differ only in the convergence criterion. The optimization algorithm was "Adam". I know that these algorithms have an stochastic nature. However, I don't understand why when I remove the convergence criterial from the second training process, I get the first training loss curve. If being stichastic means that we should get a different behavior every time we run the code, then why I am getting the first loss curve after removing the convergence criterion?

[Matt J]

It's not clear to me how strong the differences are. The axes limits are not the same in the 2 plots.

Also, we have not been shown the 3rd plot (on the same axes as the first two) with the training curve you get after reverting back to default settings.

[Memo Remo]

Dear Matt,

I appreciate your attention.

I used another approach to train this model for now. I will check this problem again as soon as I can and provide more information. Thank you.