Version 0.12.1

What's Changed

• Pass num_training_steps / num_warmup_steps to custom LR schedulers by @kohya-ss in #2452
• Add enable_multiline_caption: captions with line breaks and multiple caption candidates by @kohya-ss in #2454
• Apply caption_prefix / caption_suffix etc. when caching Text Encoder outputs by @kohya-ss in #2455
• README-ja: add the enable_multiline_caption entry (PR #2454) by @kohya-ss in #2456
• feat: MOSS-Image support, stage 1: model and inference by @kohya-ss in #2458
• feat: MOSS-Image support, stage 2: strategies and caching by @kohya-ss in #2459
• feat: MOSS-Image LoRA training (stage 3: trainer, generic LoRA network) by @kohya-ss in #2460
• feat: ConvRot INT8 for MOSS-Image (stage 4: DiT and text encoder, ported from Musubi Tuner) by @kohya-ss in #2461
• docs: add the MOSS-Image LoRA training guide and update README by @kohya-ss in #2462
• docs: MOSS-Image guide: 8GB GPUs are not covered yet by @kohya-ss in #2463
• docs: update README change history for version 0.12.1 by @kohya-ss in #2464
• Version 0.12.1 by @kohya-ss in #2465

Full Changelog: v0.12.0...v0.12.1