FAST Dreambooth with Multiple Concepts!

Want to run Dreambooth for Stable Diffusion locally so you can train multiple concepts at once really quickly? Not a problem! Runs on Google Colab as well, so you don't actually need a modern computer to train. Works on Microsoft Windows (partly), but for the lowest VRAM usage you'll need to use Linux (as with most AI stuff). Also remember to check the table on the repository page for the various configurations and VRAM used! Works with the inpainting model too - just use the inpainting scripts instead. Enjoy! Note: Always check the repo for the latest updates / changes as they happen fast! Update: For bitsandbytes on MS Windows, see https://github.com/TimDettmers/bitsandbytes/issues/30 - you'll need to install those 2 dlls into wherever your environment is (such as X:\YourAnacondaInstall\envs\diffusers\Lib\site-packages\bitsandbytes\) and then edit the 2 files as described (cuda_setup/main.py and ./cextension.py). Basically, you just add "binary_name="libbitsandbytes_cuda116.dll" before the return in evaluate_cuda_setup(), then search and replace the other string in the cextension.py file. Links: Shivam's Diffusers Fork - https://github.com/ShivamShrirao/diffusers/tree/main/examples/dreambooth Xformers - https://github.com/facebookresearch/xformers Notes file - https://pastebin.com/7Jx6VhgA Example launch.bat - https://pastebin.com/MUd79UWZ Anaconda for Windows Beginners - https://youtu.be/OjOn0Q_U8cY How do you do that avatar thing - https://youtu.be/Z7TLukqckR0 Sintel: https://durian.blender.org/ © copyright Blender Foundation | durian.blender.org Elephant's dream: https://orange.blender.org/ © copyright 2006, Blender Foundation / Netherlands Media Art Institute / www.elephantsdream.org Tux image: https://commons.wikimedia.org/wiki/File:Tux.svg Attribution: [email protected] Larry Ewing and The GIMP