//=time() ?>
Textual inversion is memory-efficient and fast, but cannot reconstruct with high fidelity,
Dreambooth has high-fidelity, but heavy and overfits unless you have large set of prior preservation.
Can you take the best of both worlds, by Pivotal Tuning in low-rank space? Yes! [1/n]