How to fine tune already trained dreambooth model?

I trained a dreambooth model for 3000 steps. But the result were not good and I want to train it again for 1000 more steps. How can I go about doing it ?

Should I pass the 3000 steps output directory as --pretrained_model_name_or_path ? or should I keep using the “CompVis/stable-diffusion-v1-4” and overwrite the output directory ?