Why my model forgets all other concepts after Dreambooth training?

[Renaldas111]

I remember when training models via Dreambooth by AUTO1111 webui, my model was able to accomodate a new concept without forgetting all what it knew. Now, training the models via Dreambooth, I am able to get them to reproduce only the trained concepts, all other knowledge is lost, if I train for 1 concept, any token I enter produces only this concept, if I train for several concepts - I receive a mix of my trained concepts.
Why I lose all the previous knowledge of a model? I am training both with and without regularization images, the result is always the same.