Kohya S.
Kohya S.
Each concept can have a different repeat in DreamBooth method. But metadata .json (fine tuning method) does not have the feature. One idea is that the folder name for fine...
Currently the epoch starts from 1 when resuming. It is better to be able to set arbitrary number by argument. (Or automatically taken from the file name.)
The implementation of Textual Inversion in Diffusers is different from CompVis. So, it might be required to implement that from scratch. - https://github.com/rinongal/textual_inversion - https://github.com/huggingface/diffusers/tree/main/examples/textual_inversion
original #1250
original PR #1223
support SD3
- [x] Replace SD3Tokenizer with the original CLIP-L/G/T5 tokenizers. - [x] Extend the max token length to 256 for T5XXL. - [x] Refactor caching for latents. - [x] Refactor caching...
WIP Sample image generation is not implemented yet.
Workflow: 1. Give the source code etc. to Gemini and have them write the documentation in Japanese or English. 2. Correct errors in documentation. 3. Add translation with Claude. 4....