v21.2.4

• 2023/03/12 (v21.2.4):
  • Fix issue with kohya locon not training the convolution layers
  • Update LyCORIS module version
  • Update LyCORYS locon extract tool

v21.2.3

• 2023.03/12 (v21.2.3):
  • Add validation that all requirements are met before starting the GUI.
• 2023/03/11 (v21.2.2):
  • Add support for LoRA LoHa type. See https://github.com/KohakuBlueleaf/LyCORIS for more details.

v21.2.1

• 2023/03/11 (v21.2.2):
  • Add support for LoRA LoHa type. See https://github.com/KohakuBlueleaf/LyCORIS for more detais.

v21.2.0

• 2023/03/09 (v21.2.0):
  • Fix issue #335
  • Add option to print LoRA trainer command without executing it
  • Add support for samples during trainin via a new Sample images config accordion in the Training parameters tab.
  • Added new Additional parameters under the Advanced Configuration section of the Training parameters tab to allow for the specifications of parameters not handles by the GUI.
  • Added support for sample as a new Accordion under the Training parameters tab. More info about the prompt options can be found here: kohya-ss/sd-scripts#256 (comment)
  • There may be problems due to major changes. If you cannot revert back to the previous version when problems occur, please do not update for a while.
  • Minimum metadata (module name, dim, alpha and network_args) is recorded even with --no_metadata, issue kohya-ss/sd-scripts#254
  • train_network.py supports LoRA for Conv2d-3x3 (extended to conv2d with a kernel size not 1x1).
    • Same as a current version of LoCon. Thank you very much KohakuBlueleaf for your help!
    • LoCon will be enhanced in the future. Compatibility for future versions is not guaranteed.
    • Specify --network_args option like: --network_args "conv_dim=4" "conv_alpha=1"
    • Additional Networks extension version 0.5.0 or later is required to use 'LoRA for Conv2d-3x3' in Stable Diffusion web UI.
    • Stable Diffusion web UI built-in LoRA does not support 'LoRA for Conv2d-3x3' now. Consider carefully whether or not to use it.
  • Merging/extracting scripts also support LoRA for Conv2d-3x3.
  • Free CUDA memory after sample generation to reduce VRAM usage, issue kohya-ss/sd-scripts#260
  • Empty caption doesn't cause error now, issue kohya-ss/sd-scripts#258
  • Fix sample generation is crashing in Textual Inversion training when using templates, or if height/width is not divisible by 8.
  • Update documents (Japanese only).
  • Dependencies are updated, Please upgrade the repo.
  • Add detail dataset config feature by extra config file. Thanks to fur0ut0 for this great contribution!
    • Documentation is here (only in Japanese currently.)
    • Specify .toml file with --dataset_config option.
    • The options supported under the previous release can be used as is instead of the .toml config file.
    • There might be bugs due to the large scale of update, please report any problems if you find at https://github.com/kohya-ss/sd-scripts/issues.
  • Add feature to generate sample images in the middle of training for each training scripts.
    • --sample_every_n_steps and --sample_every_n_epochs options: frequency to generate.
    • --sample_prompts option: the file contains prompts (each line generates one image.)
    • The prompt is subset of gen_img_diffusers.py. The prompt options w, h, d, l, s, n are supported.
    • --sample_sampler option: sampler (scheduler) for generating, such as ddim or k_euler. See help for useable samplers.
  • Add --tokenizer_cache_dir to each training and generation scripts to cache Tokenizer locally from Diffusers.
    • Scripts will support offline training/generation after caching.
  • Support letents upscaling for highres. fix, and VAE batch size in gen_img_diffusers.py (no documentation yet.)

v21.1.5

• 2023/03/05 (v21.1.5):
  • Add replace underscore with space option to WD14 captioning. Thanks @sALTaccount!
  • Improve how custom preset is set and handles.
  • Add support for --listen argument. This allow gradio to listen for connections from other devices on the network (or internet). For example: gui.ps1 --listen "0.0.0.0" will allow anyone to connect to the gradio webui.
  • Updated Resize LoRA tab to support LoCon resizing. Added new resize

v21.1.4

• 2023/03/05 (v21.1.4):
  • Removing legacy and confusing use 8bit adam chackbox. It is now configured using the Optimiser drop down list. It will be set properly based on legacy config files.

v21.1.3

• 2023/03/04 (v21.1.3):
  • Fix progress bar being displayed when not required.
  • Add support for linux, thank you @devNegative-asm
• 2023/03/03 (v21.1.2):
  • Fix issue #277
  • Fix issue #278 introduce by LoCon project switching to pip module. Make sure to run upgrade.ps1 to install the latest pip requirements for LoCon support.

v21.1.1b

Bug Fixe release:

Upgrade Gradio release
Fix bugs related to new gragio release
Fix bad version of LoCon gui

v21.1.1

• 2023/03/02 (v21.1.1):
  • Emergency fix for #261

v21.1.0

• 2023/03/02 (v21.1.0):
  • Add LoCon support (https://github.com/KohakuBlueleaf/LoCon.git) to the Dreambooth LoRA tab. This will allow to create a new type of LoRA that include conv layers as part of the LoRA... hence the name LoCon. LoCon will work with the native Auto1111 implementation of LoRA. If you want to use it with the Kohya_ss additionalNetwork you will need to install this other extension... until Kohya_ss support it nativelly: https://github.com/KohakuBlueleaf/a1111-sd-webui-locon