sd_dreambooth_extension VS lora

I am trying to use StableDiffusion via AUTOMATIC1111 with the Dreambooth extension

When I first attempted SD training, I was very frustrated. It wasn't until I found this obscure forum thread on Github that I actually started producing great results with Dreambooth. Because I have such satisfactory results, I'm very reluctant to beat my brains against LoRa and its related training techniques. I gave up trying to train TI embeddings a long time ago. And I never figured out how to train or how to use hypernetworks. I've only been able to get good results with Dreambooth directly because of that thread I linked above. I make LoRas by extracting them from Dreambooth-trained checkpoints. And I have no idea if I'm doing the extractions the right way or not.

Getting the same with automatic1111 and sd_dreambooth extension. Check out more here in the issues log: https://github.com/d8ahazard/sd_dreambooth_extension/issues/1266

It's several moths old and many things have changed. But the spreadsheet available through this thread on Github has been indispensable for me when I train Dreambooth models. I'm astounded no one talks about it. I bring it up all the time. The research presented there should be continued. I'd love to see similar research done for SD v2.1.

Training rate is paramount. Read this Github thread.

https://github.com/d8ahazard/sd_dreambooth_extension/discussions/547/ (in depth training principles understanding)

sd_dreambooth_extension https://github.com/d8ahazard/sd_dreambooth_extension.git main 926ae204 Fri Mar 31 15:12:45 2023 unknown

also the Dreambooth extension supports API (https://github.com/d8ahazard/sd_dreambooth_extension/blob/main/scripts/api.py) so i'm not sure where do you get those news :/

Diffusion unet has an "extended" version nowadays that applies to the resnet part as well as the cross-attention: https://github.com/cloneofsimo/lora

Absolutely. But that doesn't matter because you only have to train it at scale, once. There are papers released already that show it's possible to update weights in small sections. You won't have to wait for the next monolithic LLM to drop to get up to date information. It will start to learn in bits and pieces.

No, it's a deep learning thing

Low Rank Adaptation of Large Language Models.

Recently, I have seen the LoRA technique (Low-Rank Adaptation of Large Language Models) as a popular method for fine-tuning LLMs and other models.

Enter LoRA, a method proposed for adapting pre-trained models to specific tasks[2]. By freezing pre-trained model weights and injecting trainable rank decomposition matrices into the transformer architecture, LoRA can reduce the number of trainable parameters and the GPU memory requirement, making the adaptation of LLMs for downstream tasks more feasible.

Open-source LLM projects have largely solved this using Low-Rank Adaptation of Large Language Models (LoRA): https://arxiv.org/abs/2106.09685

Apparently an RTX 4090 running overnight is sufficient to produce a fine-tuned model that can spit out new Harry Potter stories, or whatever...

Alpaca was trained for $600 ($100 for the smaller model) and offers outputs competitive with ChatGTP. https://arxiv.org/abs/2106.09685