LECO is able to make a Lora from a prompt that you want to boost/erase.

A textual inversion is equivalent to some tokens, so ... it should be possible to make a Lora out of a text inversion.

I made a Lora version of my B-Horror text inversion, training source included.

NB: if you use batching, make sure to remove the zeroing of gradients at every loop.