Non convergence of Loss function

[DuXikb]

Is this model a problem of non convergence of Loss function with training process? Do I need to change the hyperparameters when I use the training set that comes with the project and does not converge?