ai-toolkit

ostris/ai-toolkit

Fork 0

mirror of https://github.com/ostris/ai-toolkit.git synced 2026-01-31 02:49:46 +00:00

Commit Graph

Select branches

Hide Pull Requests

WIP

dev

development

kohya-sdxl

lumina2

main

multi-gpu

qwen_image_edit

sdxl

wavelet_loss

#107

#109

#115

#12

#120

#124

#125

#128

#128

#138

#138

#141

#143

#155

#155

#156

#156

#158

#158

#160

#160

#168

#168

#173

#176

#177

#179

#179

#183

#184

#184

#185

#185

#191

#191

#196

#224

#224

#226

#226

#232

#233

#233

#244

#244

#245

#245

#246

#246

#249

#25

#25

#251

#251

#253

#253

#256

#263

#263

#264

#265

#265

#271

#273

#273

#274

#275

#275

#278

#283

#283

#290

#290

#300

#300

#302

#302

#303

#303

#304

#307

#318

#318

#326

#326

#331

#331

#333

#333

#341

#341

#342

#342

#343

#350

#350

#353

#353

#354

#354

#359

#362

#362

#373

#373

#375

#375

#377

#381

#381

#382

#382

#383

#385

#385

#392

#392

#395

#407

#407

#41

#417

#418

#420

#420

#422

#426

#427

#427

#430

#430

#431

#431

#432

#432

#434

#442

#442

#446

#447

#452

#455

#459

#459

#461

#461

#464

#465

#465

#467

#469

#469

#473

#473

#474

#488

#488

#489

#489

#492

#492

#494

#494

#496

#498

#498

#499

#499

#500

#509

#511

#513

#519

#519

#520

#520

#521

#522

#523

#525

#527

#527

#540

#547

#547

#551

#551

#555

#558

#559

#559

#563

#563

#564

#564

#567

#578

#578

#580

#580

#581

#581

#583

#594

#599

#60

#600

#600

#601

#601

#606

#606

#608

#608

#61

#610

#611

#614

#614

#616

#616

#629

#629

#630

#630

#633

#633

#637

#637

#639

#639

#644

#646

#648

#649

#649

#655

#655

#658

#658

#660

#661

#661

#666

#666

#668

#668

#673

#673

#682

#685

#685

#687

#687

#69

#69

#7

#80

#86

#86

#9

#95

1069dee0e4 Added ui sopport for multi control samples and datasets. Added qwen image edit 5209 to the ui Jaret Burkett 2025-09-25 11:10:02 -06:00
454be0958a Initial support for qwen image edit plus Jaret Burkett 2025-09-24 11:39:10 -06:00
f74475161e Add stepped loss type Jaret Burkett 2025-09-22 15:50:12 -06:00
28728a1e92 Added experimental dfe 5 Jaret Burkett 2025-09-21 10:48:52 -06:00
20dfe1b4d5 Small double tap of detach on qwen just for good measure Jaret Burkett 2025-09-18 08:22:04 -06:00
390e21bec6 Integrate dataset level trigger words and allow them to be cached. Default to global trigger if it is set. Jaret Burkett 2025-09-18 03:29:18 -06:00
3cdf50cbfc Merge pull request #426 from squewel/prior_reg Jaret Burkett 2025-09-18 03:03:18 -06:00
e27e229b36 add prior_reg flag to FileItemDTO squewel 2025-09-18 02:09:39 +03:00
e4ae97e790 add dataset-level distillation-style regularization max 2025-09-18 01:11:19 +03:00
2120dc5936 Upgrade job to new ui trainer to fix issue with slider config showing up on old configs. Jaret Burkett 2025-09-17 13:41:48 -06:00
24a576ad07 Regularize the slider targets. Jaret Burkett 2025-09-17 09:36:33 -06:00
218f673e3d Added support for new concept slider training script to CLI and UI Jaret Burkett 2025-09-16 10:22:34 -06:00
3666b112a8 DEF for fake vae and adjust scaling Jaret Burkett 2025-09-12 18:09:08 -06:00
b95c17dc17 Add initial support for chroma radiance Jaret Burkett 2025-09-10 08:41:05 -06:00
af6fdaaaf9 Add ability to train a full rank LoRA. (experimental) Jaret Burkett 2025-09-09 07:36:25 -06:00
645046701b Comment out fast stop watcher. Could potentiallty be causing some weird issues. Need to investigate. Jaret Burkett 2025-09-04 08:26:57 -06:00
f699f4be5f Add ability to set transparent color for control images Jaret Burkett 2025-09-02 11:08:44 -06:00
85dcae6e2b Set full size control images to default true Jaret Burkett 2025-09-02 10:30:42 -06:00
7040d8d73b Preperation for audio Jaret Burkett 2025-09-02 07:26:50 -06:00
0f2239ca23 Add force sample toggle to the ui Jaret Burkett 2025-08-31 16:58:27 -06:00
193c1b2dfa Add a watcher to constantly check for stop signal from the UI. This will force a stop within 2 seconds instead of having to wait on a long hung process. Jaret Burkett 2025-08-31 16:58:01 -06:00
6fc9ec1396 Added example config for training wan22 14b 24GB on images Jaret Burkett 2025-08-28 13:08:49 -06:00
056711d4ed Fix issue with wan22 14b that woudl load both transformers temporarily resulting in oom on 24GB. Jaret Burkett 2025-08-28 13:06:31 -06:00
e3349414fd Updated runpod docs Jaret Burkett 2025-08-28 11:40:48 -06:00
9ef425a1c5 Fixed issue with training qwen with cached text embeds with a batch size more than 1 Jaret Burkett 2025-08-28 08:07:12 -06:00
fc5b41666a Switch order to save first, then sample. Jaret Burkett 2025-08-27 11:07:03 -06:00
1f541bc5d8 Changes to handle a different DFE arch Jaret Burkett 2025-08-27 11:05:16 -06:00
fd13bd73a6 Add a Download button on samples to download all the samples as a zip file Jaret Burkett 2025-08-27 09:12:46 -06:00
5ad190b11d Improve UI for sample images when there are no samples Jaret Burkett 2025-08-27 08:10:24 -06:00
d0338b8b0b Allow dropping images directly into dataset folder without having to open the add images modal. Improve ui flow of dataset messaging. Jaret Burkett 2025-08-25 14:04:48 -06:00
37eda7b2e2 Add a tab to the UI to show the config file for the job. Read only. Jaret Burkett 2025-08-25 13:08:40 -06:00
119653c3f2 Force width, height, and num frames to always be the proper sizes for Wan 2.2 models Jaret Burkett 2025-08-25 10:33:28 -06:00
ea01a1c7d0 Fixed a bug where samples would fail if merging in lora on sampling for unquantized models. Quantize non ARA modules as uint8 when using an ARA Jaret Burkett 2025-08-25 09:21:40 -06:00
f48d21caee Upgrade a LoRA rank if the new one is larger so users can increase the rank on an exiting training job and continue training at a higher rank. Jaret Burkett 2025-08-24 13:40:25 -06:00
24372b5e35 Add toggles to the UI to add flipped versions of the datasets, X, Y or both. Jaret Burkett 2025-08-24 13:39:04 -06:00
5c27f89af5 Add example config for qwen image edit Jaret Burkett 2025-08-23 18:20:36 -06:00
554dfb33bc Added example config file for qwen image at 24GB Jaret Burkett 2025-08-23 12:37:46 -06:00
823e690703 Changed auth to use the wording 'password' instead of 'token' and give information about defaults and how to change the password. Jaret Burkett 2025-08-23 09:30:44 -06:00
e1fd411665 Added support for Chroma1 official release. Will still use single file verstion instead of the diffusers version. Jaret Burkett 2025-08-23 09:06:28 -06:00
0d6d027248 Update supporters info Jaret Burkett 2025-08-23 08:58:31 -06:00
b6f43fb7c2 Merge pull request #383 from ostris/qwen_image_edit Jaret Burkett 2025-08-22 10:29:25 -06:00
59ff4efae5 Add support for training Qwen Image Edit in the UI qwen_image_edit Jaret Burkett 2025-08-22 10:26:46 -06:00
aa99784b89 Add control to prompot encodings in the trainer when not cached Jaret Burkett 2025-08-21 16:52:13 -06:00
bf2700f7be Initial support for finetuning qwen image. Will only work with caching for now, need to add controls everywhere. Jaret Burkett 2025-08-21 16:41:17 -06:00
38d3814be7 Added 4bit ARAs for Wan 2.2 14b models Jaret Burkett 2025-08-21 08:16:07 -06:00
83deaec417 Minor bug fixes Jaret Burkett 2025-08-21 08:05:34 -06:00
d2bbe1872c Add support for fine tuning Wan 2.2 I2V 14B Jaret Burkett 2025-08-18 11:43:32 -06:00
b3e666daf4 Fix issue with wan22 14b where timesteps were generated not in the current boundary. Jaret Burkett 2025-08-16 21:16:48 -06:00
6fffadfc0e Fixed a bug that prevented training just one stage of Wan 2.2 14b Jaret Burkett 2025-08-16 18:07:21 -06:00
280aca685f Merge pull request #377 from ostris/wan22_14b Jaret Burkett 2025-08-16 14:25:23 -06:00
1029fa8743 version bump Jaret Burkett 2025-08-16 13:39:40 -06:00
8ea2cf00f6 Added training to the ui. Still testing, but everything seems to be working. Jaret Burkett 2025-08-16 05:51:37 -06:00
ca7bfa414b Increase max number of samples to 40 Jaret Burkett 2025-08-16 05:27:38 -06:00
1c96b95617 Fix issue where sometimes the transformer does not get loaded properly. Jaret Burkett 2025-08-14 14:24:41 -06:00
3413fa537f Wan22 14b training is working, still need tons of testing and some bug fixes Jaret Burkett 2025-08-14 13:03:27 -06:00
be71cc75ce Switch to unified text encoder for wan models. Pred for 2.2 14b Jaret Burkett 2025-08-14 10:07:18 -06:00
e12bb21780 Quantize blocks sequentialls without a ARA Jaret Burkett 2025-08-14 09:59:58 -06:00
3ff4430e84 Fix issue with fake text encoder unload Jaret Burkett 2025-08-14 09:33:44 -06:00
5501521c9f Link to easy install script Jaret Burkett 2025-08-13 12:26:10 -06:00
85bad57df3 Fix bug that would use EMA when set false Jaret Burkett 2025-08-13 11:39:40 -06:00
259d68d440 Added a flushg during sampling to prevent spikes on low vram qwen Jaret Burkett 2025-08-12 12:57:18 -06:00
69ee99b6e1 Fix issue with base model version Jaret Burkett 2025-08-12 09:26:48 -06:00
77b10d884d Add support for training with an accuracy recovery adapter with qwen image Jaret Burkett 2025-08-12 08:21:36 -06:00
4ad18f3d00 Clip max token embeddings to the max rope length for qwen image to solve for an error for super long captions > 1024 Jaret Burkett 2025-08-10 08:44:41 -06:00
f0105c33a7 Fixed issue that sometimes happens in qwen image where text seq length is wrong Jaret Burkett 2025-08-09 16:33:05 -06:00
ccd449ec49 Update supporters Jaret Burkett 2025-08-08 11:04:45 -06:00
bb6db3d635 Added support for caching text embeddings. This is just initial support and will probably fail for some models. Still needs to be ompimized Jaret Burkett 2025-08-07 10:27:55 -06:00
4c4a10d439 Remove vision model from qwen text encoder as it is not needed for image generation currently Jaret Burkett 2025-08-06 11:40:02 -06:00
14ccf2f3ce Refactor qwen5b model code to be qwen 5b specific Jaret Burkett 2025-08-06 10:54:56 -06:00
5d8922fca2 Add ability to designate a dataset as i2v or t2v for models that support it Jaret Burkett 2025-08-06 09:29:47 -06:00
1755e58dd9 Update generation script to handle latest models. Jaret Burkett 2025-08-05 08:55:16 -06:00
6bb3aed9a2 Merge pull request #359 from ostris/qwen_image Jaret Burkett 2025-08-04 15:51:01 -06:00
74b4d2d291 Version bump Jaret Burkett 2025-08-04 15:49:32 -06:00
23327d5659 Add qwen image to the ui Jaret Burkett 2025-08-04 15:48:51 -06:00
93202c7a2b Training working for Qwen Image Jaret Burkett 2025-08-04 21:14:30 +00:00
9da8b5408e Initial but untested support for qwen_image Jaret Burkett 2025-08-04 13:29:37 -06:00
9dfb614755 Initial work for training wan first and last frame Jaret Burkett 2025-08-04 11:37:26 -06:00
ef1d60ba34 Update wan 2.2 5b timestep distribution to weighted. Jaret Burkett 2025-07-30 10:13:22 -06:00
75f688766d Version bump Jaret Burkett 2025-07-29 09:30:54 -06:00
a558d5b68f Move transformer back to device on aggresive wan 2.2 pipeline after generation. Jaret Burkett 2025-07-29 09:13:47 -06:00
1d1199b15b Fix bug that prevented training wan 2.2 with batch size greater than 1 Jaret Burkett 2025-07-29 09:06:25 -06:00
f453e28ea3 Fixed deprecation of lumina pipeline error Jaret Burkett 2025-07-29 08:26:51 -06:00
ca7c5c950b Add support for Wan2.2 5B Jaret Burkett 2025-07-29 05:31:54 -06:00
e55116d8c9 Added hidream low vram options Jaret Burkett 2025-07-27 18:29:46 -06:00
99705ec8be Add support in UI for Hidream E1 Jaret Burkett 2025-07-27 18:13:36 -06:00
ed8d14225f Add ability to set the quantization type for text encoders and transformer in the ui Jaret Burkett 2025-07-27 18:00:53 -06:00
b717586ee2 Version bump Jaret Burkett 2025-07-27 15:13:28 -06:00
cefa2ca5fe Added initial support for Hidream E1 training Jaret Burkett 2025-07-27 15:12:56 -06:00
3f518d9951 Add sharpening before losses with a split loss on vae training Jaret Burkett 2025-07-27 15:11:56 -06:00
77dc38a574 Some work on caching text embeddings Jaret Burkett 2025-07-26 09:22:04 -06:00
0d89c44624 Bug fixes on vae trainer. Allow to target params for vae training. Jaret Burkett 2025-07-26 09:20:22 -06:00
3e14a674ac Fix upload progress for datasets in the ui Jaret Burkett 2025-07-26 09:07:30 -06:00
523c159579 Add vram flag to some models in the ui Jaret Burkett 2025-07-24 07:02:46 -06:00
c5eb763342 Improvements to VAE trainer. Allow CLIP loss. Jaret Burkett 2025-07-24 06:50:56 -06:00
ca5cf827a1 Version bump Jaret Burkett 2025-07-20 12:20:46 -06:00
b1bff66d52 Merge pull request #343 from davertor/fix_kontext_bs Jaret Burkett 2025-07-20 12:00:55 -06:00
a77ba5a089 fix: Guidance incorrect shape Daniel Verdu 2025-07-18 12:49:18 +02:00
8610c6ed7f Made it easy to add control images to the samples in the UI dev Jaret Burkett 2025-07-17 12:00:48 -06:00
e25d2feddf Use scale shift in vae latent space for vae trainer Jaret Burkett 2025-07-17 08:14:07 -06:00
f500b9f240 Add ability to do more advanced sample prompt objects to prepart for a UI rework on control images and other things. Jaret Burkett 2025-07-17 07:13:35 -06:00