2 Colab adaptations for both hlky and AUTOMATIC1111 Webuis of stable diffusion, implementing the optimization suggested by https://github.com/MatthieuTPHR : huggingface/diffusers#532, using the MemoryEfficientAttention implementation from xformers (cc. @fmassa, @danthe3rd, @blefaudeux) to both speedup the cross-attention speed and decrease its GPU memory requirements.
All you have to do is enter your huggingface token only once and you're all set, the colabs will install the repos and the models inside Gdrive, so the loading will be fast everytime you use it, enjoy !!
AUTOMATIC1111 HLKY Relaxed mode![](https://github.com/TheLastBen/fast-stable-diffusion/raw/main/Dreambooth/1.jpg)
![](https://github.com/TheLastBen/fast-stable-diffusion/raw/main/Dreambooth/2.jpg)
![](https://github.com/TheLastBen/fast-stable-diffusion/raw/main/Dreambooth/3.jpg)
If you encounter any issue or you want to update to latest webui version, remove the folder "sd" or "stable-diffusion-webui" from your GDrive (and GDrive trash) and rerun the colab.
Train your model using this easy simple and fast colab, all you have to do is enter you huggingface token once, and it will cache all the files in GDrive, including the trained model and you will be able to use it directly from the colab, make sure you use high quality reference pictures for the training. Thanks to https://github.com/jachiam for his diffusers to CKPT script, the trained models now can now be used in SD WEBUI from AUTOMATIC1111 and hlky and others.
DreamBooth
TODO :
- Paperspace support
- Deepspeed