Train dreambooth from textual inversion checkpoint

following the blog Training Stable Diffusion with Dreambooth using Diffusers at the very end is mentioned you trained dreambooth from a textual inversion checkpoint.
How can I do that?
If I try to use --resume_from_checkpoint from the textual inversion checkpoint path it complains it doesn’t find the text_encoder config.json
what should be the args fro the to do this?
thank you