Version 0.2.0

Features

• Add LoRA support: trained LoRA models can be used with the extension for 1111's web UI.
• Make BLIP captioning not to require extra downloading.
• Fix bugs.

What's Changed

• Proposed file structure rework and required file changes by @bmaltais in #1
• Make BLIP working without extra downloading. etc. by @kohya-ss in #2
• Merge requirements by @bmaltais in #3
• Fix relative path to train dir is not working. etc. by @kohya-ss in #7
• Fix text encoder training is not stopped #4 by @kohya-ss in #8
• Fix conversion error for v1 Diffusers->ckpt. #10 by @kohya-ss in #11
• Fix npz file name for images with dots #12 by @kohya-ss in #13
• Add LoRA training/generating. by @kohya-ss in #16
• set default dataset_repeats to 1, resolution check by @kohya-ss in #22
• Fix typos by @shirayu in #25
• rename another position_ids key (supports wd v1.4) by @kohya-ss in #29

New Contributors

• @bmaltais made their first contribution in #1
• @kohya-ss made their first contribution in #2

Full Changelog: v0.1.0...v0.2.0

First release.

Same as:
train_db_fixed: v18
diffusers_fine_tunining: v9
gen_img_diffusers: v9