Linoy Tsaban

[Advanced DreamBooth LoRA SDXL] Support EDM-style training (follow up of #7126)

2

This PR adds support for EDM-style training to the advanced script, based off PR #7126

WIP

[Dreambooth LoRA training] Add `edm` tags + scheduler type to model card

3

- Add `edm-training` tag to canonical and advanced dreambooth lora training scripts when EDM style training is used - Add scheduler type to model card for inference with EDM schedulers

Add B-Lora training option to the advanced dreambooth lora script

4

Adds an option to train a dreambooth lora targeting using a single ref image + targeting specific unet blocks only, subsequently allowing for easy combination of style/content LoRAs, as proposed...

add accelerate notebook_launcher to sd-xl dreambooth

5

adds accelerate notebook_launcher to improve user experience + some minor ui improvements building on top of #424 training script was copied from train_dreambooth_lora_sdxl.py cc @sayakpaul @apolinario 🤗

sd-xl dreambooth QOL improvements

12

A few QOL improvements: - removed assert from image_grid - improved login experience - add link to model at the end of training - added local image upload option

[advanced dreambooth lora] add clip_skip arg

1

adds `--clip_skip` to the advanced sdxl dreambooth lora trainer

[SD3-5 dreambooth lora] update model cards

1

small updates to model card saving

[flux dreambooth lora training] make LoRA target modules configurable + small bug fix

1

new feature for the Flux dreambooth lora training script: - make LoRA target modules configurable through `--lora_blocks` - change the current default target modules to not be attention layers only...

[SD 3.5 Dreambooth LoRA] support configurable training block & layers

2

This PR adds - `--lora_layers` and `--lora_blocks` to dreambooth lora training script to allow targeting specific blocks & layers. - This is generally proven to be useful, and specifically for...