How to tune hyperparameter, where data is passing through model train function step by step?

I am training a ML model, where instead of pushing the data in a whole, I wanted to give data step by step.

So, like saving weights in Deep learning model. Can we save parameters of some part of training data and then again load these parameters to further tune the hyper-parameters?

