Why the size of model vary with different number of workers?

I 'm training a fm model on hadoop right now, and I notice the number of parts(files) of the model is equal to the number of worker. But I notice that the size of each part remain same when I change the worker number, that means the size of model also changed when I change number of worker. Could you pls give me some hint about this question? 