- Feature: Removed a lot of dependencies and updated the model to use the latest version of the model
- Model Update: replaced Florence with a new model, MiaoshouAI/Florence-2-base-PromptGen-v2.0. This model is fine-tune of Florence2
- Hotfix: Update caption mode for JoyCaption, fix some typoes
- Caption Features: Added JoyCaption Alpha to options for captioning. This is a new feature that allows for richer captions to be generated.
- Caption UI: Added a new gradio UI for editing captions. This allows for more control over the captioning process.
- Caption Features: New caption options and the abailty to install flash-attn from the command line.
- Dateset Features: Now allows for any popular dataset file to be exported for training. This includes
JSONL
,HFJSON
,CSV
, andJSON
.
- Images Features: Now generates two sets of images at 512, 1024. training is done on a 672x672 image.
- Initial Release: Established the initial codebase for the project, laying the foundation for future development and enhancements.