Dreambooth

[JohnTigue]

Dreambooth continues to get used in innovative projects. Invoke2.2 UI mentions that Dreambooth will be usable within Invoke. So, Dreambooth is on the ToDo list.

DreamBooth is an AI algorithm that allows you to teach new subjects or even styles to existing Stable Diffusion models very successfully, such as teaching the face of a person.

[JohnTigue]

The Dreambooth paper: DreamBooth: Fine Tuning Text-to-Image Diffusion Models for Subject-Driven Generation, 202208

[JohnTigue]

Reddit: Creating a specific character via Dreambooth

[JohnTigue]

Reddit:

I cant say since dreambooth requires +12GB gpu, I have zero knowledge of it at the moment

[JohnTigue]

YouTube (P11m): Dreambooth & SD on Colab

[JohnTigue]

This seems to be popular (e.g., 3.5k GitHub stars, "is a phenomenal (accessible) start."): https://github.com/TheLastBen/fast-stable-diffusion

[JohnTigue]

This is essentially a 19 minute non-technical intro to DreamBooth (with cheezy embedded commercials): Is This The Death of VFX?

[JohnTigue]

DreamBooth is often compared to textual inversion (#2). The former is considered more powerul.

[JohnTigue]

Zero To Hero Stable Diffusion DreamBooth Tutorial By Using Automatic1111 Web UI - Ultra Detailed (YouTube, 1h43m)

[JohnTigue]

DreamBooth Got Buffed - 22 January Update - Much Better Success Train Stable Diffusion Models Web UI

[JohnTigue]

Seems the main branch of DreamBooth A1111 extension has been broken for weeks. Here's how to work around that, DREAMBOOTH: 10 MINS TRAINING In Stable Diffusion! AMAZING RESULTS!.

[JohnTigue]

Here's runpod's Dockerfile for fast-stable-diffusion. That might be the easiest way to get DreamBooth running inside Hynowerk. [TODO:JFT] not sure about the license.

[JohnTigue]

Looks like DreamBooth's problem may simply be that diffuers is not installed?

Exception importing api
Traceback (most recent call last):
  File "/stable-diffusion-webui/extensions/sd_dreambooth_extension/scripts/api.py", line 27, in <module>
    from dreambooth.dataclasses.db_config import from_file, DreamboothConfig
  File "/stable-diffusion-webui/extensions/sd_dreambooth_extension/dreambooth/dataclasses/db_config.py", line 10, in <module>
    from dreambooth.utils.image_utils import get_scheduler_names  # noqa
  File "/stable-diffusion-webui/extensions/sd_dreambooth_extension/dreambooth/utils/image_utils.py", line 11, in <module>
    from diffusers.schedulers import KarrasDiffusionSchedulers
ModuleNotFoundError: No module named 'diffusers'
Error loading script: main.py
Traceback (most recent call last):
  File "/stable-diffusion-webui/modules/scripts.py", line 248, in load_scripts
    script_module = script_loading.load_module(scriptfile.path)
  File "/stable-diffusion-webui/modules/script_loading.py", line 11, in load_module
    module_spec.loader.exec_module(module)
  File "<frozen importlib._bootstrap_external>", line 883, in exec_module
  File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed
  File "/stable-diffusion-webui/extensions/sd_dreambooth_extension/scripts/main.py", line 7, in <module>
    from dreambooth.dataclasses.db_config import (
  File "/stable-diffusion-webui/extensions/sd_dreambooth_extension/dreambooth/dataclasses/db_config.py", line 10, in <module>
    from dreambooth.utils.image_utils import get_scheduler_names  # noqa
  File "/stable-diffusion-webui/extensions/sd_dreambooth_extension/dreambooth/utils/image_utils.py", line 11, in <module>
    from diffusers.schedulers import KarrasDiffusionSchedulers
ModuleNotFoundError: No module named 'diffusers'

[JohnTigue]

A (currently open) InvokeAI GitHub Issue: Dreambooth support (#995)

[JohnTigue]

The Discord for the Automatic1111 DreamBooth extension: https://discord.gg/vEQtE2Rcxq

[JohnTigue]

Might as well go with the recently released update to the DreamBooth A1111 extension: How To Install New DreamBooth & Torch 2 On Automatic1111 Web UI PC For Epic Performance Gains Guide.

[JohnTigue]

I guess we'll be going for PyTorch v2 as well (multiple projects seem to be in the process of deploying it):

In this video, I will also explain how you can install Torch version 2. Torch version 2 is supposed to increase image speed generation and training speed significantly

[JohnTigue]

This Dr. Furkan Gözükara, a PhD Computer Engineer, is putting out great content on his YouTube channel, SECourses. In the context of DreamBooth, we have:

• Zero To Hero Stable Diffusion DreamBooth Tutorial By Using Automatic1111 Web UI - Ultra Detailed, 1h42m
• How To Install New DreamBooth Extension On RunPod - Automatic1111 Web UI - Stable Diffusion, 5m
• How To Install New DreamBooth & Torch 2 On Automatic1111 Web UI PC For Epic Performance Gains Guide, 9m
• Epic Web UI DreamBooth Update - New Best Settings - 10 Stable Diffusion Training Compared on RunPods 1h01m
• DreamBooth Got Buffed - 22 January Update - Much Better Success Train Stable Diffusion Models Web UI, 12m
• Automatic1111 Stable Diffusion DreamBooth Guide: Optimal Classification Images Count Comparison Test, 31m
• Training Midjourney Level Style And Yourself Into The SD 1.5 Model via DreamBooth Stable Diffusion, 19m
• 8GB LoRA Training - Fix CUDA & xformers For DreamBooth and Textual Inversion in Automatic1111 SD UI, 17m
• RunPod Fix For DreamBooth & xFormers - How To Use Automatic1111 Web UI Stable Diffusion on RunPod , 12m
• How to Use SD 2.1 & Custom Models on Google Colab for Training with Dreambooth & Image Generation , 8m

RunPod uses Docker containers…

[JohnTigue]

Here we go! It looks like diffusers versioning is a common problem with the A1111 Dreambooth extension, straight from the horse's mouth: Dreambooth Extension for Stable-Diffusion-WebUI: Installation.