Commit Graph

  • fd68703f37 Add New lr scheduler (#1393) 青龍聖者@bdsqlsz 2024-09-11 20:25:45 +08:00
  • 65b8a064f6 update README Kohya S 2024-09-10 21:20:38 +09:00
  • d10ff62a78 support individual LR for CLIP-L/T5XXL Kohya S 2024-09-10 20:32:09 +09:00
  • d29af146b8 add negative prompt for flux inference script Kohya S 2024-09-09 23:01:15 +09:00
  • ce144476cf Merge branch 'dev' into sd3 Kohya S 2024-09-07 10:59:22 +09:00
  • 62ec3e6424 Merge branch 'main' into dev Kohya S 2024-09-07 10:52:49 +09:00
  • de25945a93 Merge pull request #1550 from kohya-ss/dependabot/github_actions/crate-ci/typos-1.24.3 Kohya S. 2024-09-07 10:50:46 +09:00
  • 0005867ba5 update README, format code Kohya S 2024-09-07 10:45:18 +09:00
  • 16bb5699ac Merge pull request #1426 from sdbds/resize Kohya S. 2024-09-07 10:22:52 +09:00
  • 319e4d9831 Merge pull request #1433 from millie-v/sample-image-without-cuda Kohya S. 2024-09-07 10:19:55 +09:00
  • 2889108d85 feat: Add --cpu_offload_checkpointing option to LoRA training Kohya S 2024-09-05 20:58:33 +09:00
  • d9129522a6 set dtype before calling ae closes #1562 Kohya S 2024-09-05 12:20:07 +09:00
  • 90ed2dfb52 feat: Add support for merging CLIP-L and T5XXL LoRA models Kohya S 2024-09-05 08:39:29 +09:00
  • 56cb2fc885 support T5XXL LoRA, reduce peak memory usage #1560 Kohya S 2024-09-04 23:15:27 +09:00
  • b7cff0a754 update README Kohya S 2024-09-04 21:35:47 +09:00
  • b65ae9b439 T5XXL LoRA training, fp8 T5XXL support Kohya S 2024-09-04 21:33:17 +09:00
  • 6abacf04da update README Kohya S 2024-09-02 13:05:26 +09:00
  • 4f6d915d15 update help and README Kohya S 2024-09-01 19:12:29 +09:00
  • 1e30aa83b4 Merge pull request #1541 from sdbds/flux_shift Kohya S. 2024-09-01 19:00:16 +09:00
  • 92e7600cc2 Move freeze_blocks to sd3_train because it's only for sd3 Kohya S 2024-09-01 18:57:07 +09:00
  • ef510b3cb9 Sd3 freeze x_block (#1417) 青龍聖者@bdsqlsz 2024-09-01 17:41:01 +08:00
  • 928e0fc096 Merge pull request #1529 from Akegarasu/sd3 Kohya S. 2024-09-01 18:29:27 +09:00
  • 1bcf8d600b Bump crate-ci/typos from 1.19.0 to 1.24.3 dependabot[bot] 2024-09-01 01:33:04 +00:00
  • f8f5b16958 Merge pull request #1540 from kohya-ss/dependabot/pip/opencv-python-4.8.1.78 Kohya S. 2024-08-31 21:37:07 +09:00
  • 826ab5ce2e Merge pull request #1532 from nandometzger/main Kohya S. 2024-08-31 21:36:33 +09:00
  • 25c9040f4f Update flux_train_utils.py sdbds 2024-08-31 03:05:19 +08:00
  • 3a6154b7b0 Bump opencv-python from 4.7.0.68 to 4.8.1.78 dependabot[bot] 2024-08-31 06:21:16 +00:00
  • 2a3aefb4e4 Update train_util.py, bug fix Nando Metzger 2024-08-30 08:15:05 +02:00
  • 35882f8d5b fix Akegarasu 2024-08-29 23:03:43 +08:00
  • 34f2315047 fix: text_encoder_conds referenced before assignment Akegarasu 2024-08-29 22:33:37 +08:00
  • 8fdfd8c857 Update safetensors to version 0.4.4 in requirements.txt #1524 Kohya S 2024-08-29 22:26:29 +09:00
  • 8ecf0fc4bf Refactor code to ensure args.guidance_scale is always a float #1525 Kohya S 2024-08-29 22:10:57 +09:00
  • 930d709e3d Merge pull request #1525 from Akegarasu/sd3 Kohya S. 2024-08-29 22:08:57 +09:00
  • daa6ad5165 Update README.md Kohya S. 2024-08-29 21:25:30 +09:00
  • a0cfb0894c Cleaned up README Kohya S 2024-08-29 21:20:33 +09:00
  • 6c0e8a5a17 make guidance_scale keep float in args Akegarasu 2024-08-29 14:50:29 +08:00
  • a61cf73a5c update readme Kohya S 2024-08-27 21:44:10 +09:00
  • 3be712e3e0 feat: Update direct loading fp8 ckpt for LoRA training Kohya S 2024-08-27 21:40:02 +09:00
  • 0087a46e14 FLUX.1 LoRA supports CLIP-L Kohya S 2024-08-27 19:59:40 +09:00
  • 72287d39c7 feat: Add shift option to --timestep_sampling in FLUX.1 fine-tuning and LoRA training Kohya S 2024-08-25 16:01:24 +09:00
  • ea9242653c Merge branch 'dev' into sd3 Kohya S 2024-08-24 21:24:44 +09:00
  • d5c076cf90 update readme Kohya S 2024-08-24 21:21:39 +09:00
  • 4ca29edbff Merge pull request #1505 from liesened/patch-2 Kohya S. 2024-08-24 21:16:53 +09:00
  • 5639c2adc0 fix typo Kohya S 2024-08-24 16:37:49 +09:00
  • cf689e7aa6 feat: Add option to split projection layers and apply LoRA Kohya S 2024-08-24 16:35:43 +09:00
  • 2e89cd2cc6 Fix issue with attention mask not being applied in single blocks Kohya S 2024-08-24 12:39:54 +09:00
  • 1e8108fec9 Handle args.v_parameterization properly for MinSNR and changed prediction target liesen 2024-08-24 01:38:17 +03:00
  • 81411a398e speed up getting image sizes Kohya S 2024-08-22 22:02:29 +09:00
  • 99744af53a Merge branch 'dev' into sd3 Kohya S 2024-08-22 21:34:24 +09:00
  • afb971f9c3 fix SD1.5 LoRA extraction #1490 Kohya S 2024-08-22 21:33:15 +09:00
  • bf9f798985 chore: fix typos, remove debug print Kohya S 2024-08-22 19:59:38 +09:00
  • b0a980844a added a script to extract LoRA Kohya S 2024-08-22 19:57:29 +09:00
  • 2d8fa3387a Fix to remove zero pad for t5xxl output Kohya S 2024-08-22 19:56:27 +09:00
  • a4d27a232b Fix --debug_dataset to work. Kohya S 2024-08-22 19:55:31 +09:00
  • 98c91a7625 Fix bug in FLUX multi GPU training kohya-ss 2024-08-22 12:37:41 +09:00
  • e1cd19c0c0 add stochastic rounding, fix single block Kohya S 2024-08-21 21:04:10 +09:00
  • 2b07a92c8d Fix error in applying mask in Attention and add LoRA converter script Kohya S 2024-08-21 12:30:23 +09:00
  • e17c42cb0d Add BFL/Diffusers LoRA converter #1467 #1458 #1483 Kohya S 2024-08-21 12:28:45 +09:00
  • 7e459c00b2 Update T5 attention mask handling in FLUX Kohya S 2024-08-21 08:02:33 +09:00
  • 6ab48b09d8 feat: Support multi-resolution training with caching latents to disk Kohya S 2024-08-20 21:39:43 +09:00
  • 388b3b4b74 Merge pull request #1482 from kohya-ss/flux-merge-lora Kohya S. 2024-08-20 19:34:57 +09:00
  • dbed5126bd chore: formatting Kohya S 2024-08-20 19:33:47 +09:00
  • 9381332020 revert merge function add add option to use new func Kohya S 2024-08-20 19:32:26 +09:00
  • 6f6faf9b5a fix to work with ai-toolkit LoRA Kohya S 2024-08-20 19:16:25 +09:00
  • 92b1f6d968 Merge pull request #1469 from exveria1015/sd3 Kohya S. 2024-08-20 19:06:06 +09:00
  • c62c95e862 update about multi-resolution training in FLUX.1 Kohya S 2024-08-20 08:21:01 +09:00
  • 9e72be0a13 Fix debug_dataset to work Kohya S 2024-08-20 08:19:00 +09:00
  • 486fe8f70a feat: reduce memory usage and add memory efficient option for model saving Kohya S 2024-08-19 22:30:24 +09:00
  • 6e72a799c8 reduce peak VRAM usage by excluding some blocks to cuda Kohya S 2024-08-19 21:55:28 +09:00
  • d034032a5d update README fix option name Kohya S 2024-08-19 13:08:49 +09:00
  • a450488928 update readme Kohya S 2024-08-18 16:56:50 +09:00
  • ef535ec6bb add memory efficient training for FLUX.1 Kohya S 2024-08-18 16:54:18 +09:00
  • 7e688913ae fix: Flux の LoRA マージ機能を修正 exveria1015 2024-08-18 12:38:05 +09:00
  • 25f77f6ef0 fix flux fine tuning to work kohya-ss 2024-08-17 15:54:32 +09:00
  • 400955d3ea add fine tuning FLUX.1 (WIP) Kohya S 2024-08-17 15:36:18 +09:00
  • 7367584e67 fix sd3 training to work without cachine TE outputs #1465 Kohya S 2024-08-17 14:38:34 +09:00
  • e45d3f8634 add merge LoRA script Kohya S 2024-08-16 22:19:21 +09:00
  • 3921a4efda add t5xxl max token length, support schnell Kohya S 2024-08-16 17:06:05 +09:00
  • 739a8969bc Merge pull request #1461 from fireicewolf/sd3-devel Kohya S. 2024-08-16 14:15:24 +09:00
  • 08ef886bfe Fix AttributeError: 'FluxNetworkTrainer' object has no attribute 'sample_prompts_te_outputs' Move "self.sample_prompts_te_outputs = None" from Line 150 to Line 26. DukeG 2024-08-16 11:00:08 +08:00
  • 35b6cb0cd1 update for torchvision Kohya S 2024-08-15 22:07:35 +09:00
  • 8aaa1967bd fix encoding latents closes #1456 Kohya S 2024-08-15 22:07:23 +09:00
  • e2d822cad7 Merge pull request #1452 from fireicewolf/sd3-devel Kohya S. 2024-08-15 21:12:19 +09:00
  • 7db4222119 add sample image generation during training Kohya S 2024-08-14 22:15:26 +09:00
  • 9760d097b0 Fix AttributeError: 'T5EncoderModel' object has no attribute 'text_model' While loading T5 model in GPU. DukeG 2024-08-14 19:58:54 +08:00
  • 56d7651f08 add experimental split mode for FLUX Kohya S 2024-08-13 22:28:39 +09:00
  • 9711c96f96 update README kohya-ss 2024-08-13 21:03:17 +09:00
  • f5ce754bc2 Merge branch 'dev' into sd3 kohya-ss 2024-08-13 21:00:44 +09:00
  • 4cf42cc5d4 Merge branch 'sd3' of https://github.com/kohya-ss/sd-scripts into sd3 kohya-ss 2024-08-13 21:00:21 +09:00
  • 0415d200f5 update dependencies closes #1450 kohya-ss 2024-08-13 21:00:16 +09:00
  • a7d5dabde3 Update readme Kohya S 2024-08-12 17:09:19 +09:00
  • 4af36f9632 update to work interactive mode kohya-ss 2024-08-12 13:24:10 +09:00
  • 9e09a69df1 update README Kohya S 2024-08-12 08:19:45 +09:00
  • 74f91c2ff7 correct option name closes #1446 Kohya S 2024-08-11 21:54:10 +09:00
  • d25ae361d0 fix apply_t5_attn_mask to work Kohya S 2024-08-11 19:07:07 +09:00
  • 82314ac2e7 update readme for ai toolkit settings Kohya S 2024-08-11 11:14:08 +09:00
  • 8a0f12dde8 update FLUX LoRA training Kohya S 2024-08-10 23:42:05 +09:00
  • 358f13f2c9 fix alpha is ignored Kohya S 2024-08-10 14:03:59 +09:00
  • 808d2d1f48 fix typos Kohya S 2024-08-09 23:02:51 +09:00
  • 36b2e6fc28 add FLUX.1 LoRA training Kohya S 2024-08-09 22:56:48 +09:00