Commit Graph

  • bee8cee7e8 update README for fused optimizer Kohya S 2024-05-12 15:08:52 +09:00
  • f3d2cf22ff update README for fused optimizer Kohya S 2024-05-12 15:03:02 +09:00
  • 6dbc23cf63 Merge branch 'dev' into fused-backward-pass Kohya S 2024-05-12 14:21:56 +09:00
  • c1ba0b4356 update readme Kohya S 2024-05-12 14:21:10 +09:00
  • 607e041f3d chore: Refactor optimizer group Kohya S 2024-05-12 14:16:41 +09:00
  • 793aeb94da fix get_trainable_params in controlnet-llite training AngelBottomless 2024-05-07 18:21:31 +09:00
  • b56d5f7801 add experimental option to fuse params to optimizer groups Kohya S 2024-05-06 21:35:39 +09:00
  • 017b82ebe3 update help message for fused_backward_pass Kohya S 2024-05-06 15:05:42 +09:00
  • 2a359e0a41 Merge pull request #1259 from 2kpr/fused_backward_pass Kohya S 2024-05-06 15:01:56 +09:00
  • 3fd8cdc55d fix dylora loraplus Kohya S 2024-05-06 14:03:19 +09:00
  • c1ef6dcabc fix to work with wrapped optimizer by accelerate Kohya S 2024-05-06 13:17:14 +09:00
  • 7fe81502d0 update loraplus on dylora/lofa_fa Kohya S 2024-05-06 11:09:32 +09:00
  • 5fe9ded188 simplify codes for schedule free optimizer Kohya S 2024-05-04 21:03:47 +09:00
  • c68712635c Support new optimizer Schedule free (#1250) 青龍聖者@bdsqlsz 2024-05-04 17:56:27 +08:00
  • 52e64c69cf add debug log Kohya S 2024-05-04 18:43:52 +09:00
  • 58c2d856ae support block dim/lr for sdxl Kohya S 2024-05-03 22:18:20 +09:00
  • 8db0cadcee Add caption_separator to output for subset Dave Lage 2024-05-02 18:08:28 -04:00
  • dbb7bb288e Fix caption_separator missing in subset schema Dave Lage 2024-05-02 17:39:35 -04:00
  • 969f82ab47 move loraplus args from args to network_args, simplify log lr desc Kohya S 2024-04-29 20:04:25 +09:00
  • 834445a1d6 Merge pull request #1233 from rockerBOO/lora-plus Kohya S 2024-04-29 18:05:12 +09:00
  • fdbb03c360 removed unnecessary torch import on line 115 frodo821 2024-04-23 14:29:05 +09:00
  • 040e26ff1d Regenerate failed file If a latent file fails to load, print out the path and the error, then return false to regenerate it Cauldrath 2024-04-21 13:46:31 -04:00
  • 0540c33aca pop weights if available #1247 Kohya S 2024-04-21 17:45:29 +09:00
  • 52652cba1a disable main process check for deepspeed #1247 Kohya S 2024-04-21 17:41:32 +09:00
  • 5cb145d13b Update train_util.py 青龍聖者@bdsqlsz 2024-04-20 21:56:24 +08:00
  • b886d0a359 Cleaned typing to be in line with accelerate hyperparameters type resctrictions Maatra 2024-04-20 14:36:47 +01:00
  • 4477116a64 fix train controlnet 青龍聖者@bdsqlsz 2024-04-20 21:26:09 +08:00
  • 2c9db5d9f2 passing filtered hyperparameters to accelerate Maatra 2024-04-20 14:11:43 +01:00
  • fc374375de Allow negative learning rate This can be used to train away from a group of images you don't want As this moves the model away from a point instead of towards it, the change in the model is unbounded So, don't set it too low. -4e-7 seemed to work well. Cauldrath 2024-04-18 23:29:01 -04:00
  • feefcf256e Display name of error latent file When trying to load stored latents, if an error occurs, this change will tell you what file failed to load Currently it will just tell you that something failed without telling you which file Cauldrath 2024-04-18 23:15:36 -04:00
  • 64916a35b2 add disable_mmap to args Zovjsra 2024-04-16 16:40:08 +08:00
  • 4f203ce40d Fused backward pass 2kpr 2024-04-14 09:56:58 -05:00
  • 68467bdf4d Fix unset or invalid LR from making a param_group rockerBOO 2024-04-11 17:33:19 -04:00
  • fde8026c2d Update config_util.py gesen2egee 2024-04-11 11:29:26 +08:00
  • 89ad69b6a0 Update train_util.py gesen2egee 2024-04-11 08:42:31 +08:00
  • 459b12539b Update config_util.py gesen2egee 2024-04-11 01:52:14 +08:00
  • 3b251b758d Update config_util.py gesen2egee 2024-04-11 01:50:32 +08:00
  • 229c5a38ef Update train_util.py gesen2egee 2024-04-11 01:45:49 +08:00
  • 36d4023431 Update config_util.py gesen2egee 2024-04-11 01:39:17 +08:00
  • 086f6000f2 Merge branch 'main' into val gesen2egee 2024-04-11 01:14:46 +08:00
  • 75833e84a1 Fix default LR, Add overall LoRA+ ratio, Add log rockerBOO 2024-04-08 19:23:02 -04:00
  • faadc350a4 use skip_first_batches to skip, close pillow image resume-step-and-epoch Kohya S 2024-04-08 23:28:38 +09:00
  • 6d9338f8b5 Merge branch 'dev' into resume-step-and-epoch Kohya S 2024-04-08 12:50:18 +09:00
  • 71e2c91330 Merge pull request #1230 from kohya-ss/dependabot/github_actions/crate-ci/typos-1.19.0 Kohya S 2024-04-07 21:14:18 +09:00
  • bfb352bc43 change huber_schedule from exponential to snr v0.8.7 Kohya S 2024-04-07 21:07:52 +09:00
  • c973b29da4 update readme Kohya S 2024-04-07 20:51:52 +09:00
  • 683f3d6ab3 Merge pull request #1212 from kohya-ss/dev v0.8.6 Kohya S 2024-04-07 20:42:41 +09:00
  • dfa30790a9 update readme Kohya S 2024-04-07 20:34:26 +09:00
  • 5f0eebaa56 experimenal impl to restore step/ecoch in resuming Kohya S 2024-04-07 18:18:03 +09:00
  • d30ebb205c update readme, add metadata for network module Kohya S 2024-04-07 14:58:17 +09:00
  • 90b18795fc Add option to use Scheduled Huber Loss in all training pipelines to improve resilience to data corruption (#1228) kabachuha 2024-04-07 07:54:21 +03:00
  • 089727b5ee update readme Kohya S 2024-04-07 12:42:49 +09:00
  • 921036dd91 Merge pull request #1240 from kohya-ss/verify-command-line-args Kohya S 2024-04-07 12:27:03 +09:00
  • cd587ce62c verify command line args if wandb is enabled ykume 2024-04-05 08:23:03 +09:00
  • 1933ab4b48 Fix default_lr being applied rockerBOO 2024-04-03 12:46:34 -04:00
  • b748b48dbb fix attention couple+deep shink cause error in some reso Kohya S 2024-04-03 12:43:08 +09:00
  • c7691607ea Add LoRA-FA for LoRA+ rockerBOO 2024-04-01 15:43:04 -04:00
  • f99fe281cb Add LoRA+ support rockerBOO 2024-04-01 15:38:26 -04:00
  • 80e9f72234 Bump crate-ci/typos from 1.17.2 to 1.19.0 dependabot[bot] 2024-04-01 01:50:22 +00:00
  • 2258a1b753 add save/load hook to remove U-Net/TEs from state Kohya S 2024-03-31 15:50:35 +09:00
  • 059ee047f3 fix typo Kohya S 2024-03-30 23:02:24 +09:00
  • 2c2ca9d726 update tagger doc Kohya S 2024-03-30 22:55:56 +09:00
  • f5323e3c4b update tagger doc Kohya S 2024-03-30 22:10:37 +09:00
  • cae5aa0a56 update wd14 tagger and doc Kohya S 2024-03-30 21:48:22 +09:00
  • 6ba84288d9 Merge pull request #1216 from Disty0/dev Kohya S 2024-03-30 18:50:49 +09:00
  • 434dc408f9 update readme Kohya S 2024-03-30 17:12:36 +09:00
  • ae3f625739 Merge branch 'dev' of https://github.com/kohya-ss/sd-scripts into dev Kohya S 2024-03-30 14:57:43 +09:00
  • f1f30ab418 fix to work with num_beams>1 closes #1149 Kohya S 2024-03-30 14:57:39 +09:00
  • bc586ce190 Add --use_rating_tags and --character_tags_first for WD Tagger Disty0 2024-03-29 13:56:42 +03:00
  • 4012fd24f6 IPEX fix pin_memory Disty0 2024-03-28 21:08:16 +03:00
  • 954731d564 fix typo Disty0 2024-03-27 22:00:59 +03:00
  • dd9763be31 Rating support for WD Tagger Disty0 2024-03-27 21:53:40 +03:00
  • b86af6798d Merge pull request #1213 from Disty0/dev Kohya S 2024-03-27 23:15:33 +09:00
  • 6f7e93d5cc Add OpenVINO and ROCm ONNX Runtime for WD14 Disty0 2024-03-27 03:21:13 +03:00
  • 6c08e97e1f update readme Kohya S 2024-03-26 20:48:08 +09:00
  • 78e0a7630c Merge pull request #1206 from kohya-ss/dataset-cache Kohya S 2024-03-26 19:49:23 +09:00
  • c86e356013 Merge branch 'dev' into dataset-cache Kohya S 2024-03-26 19:43:40 +09:00
  • 5a2afb3588 Merge pull request #1207 from kohya-ss/masked-loss Kohya S 2024-03-26 19:41:31 +09:00
  • ab1e389347 Merge branch 'dev' into masked-loss Kohya S 2024-03-26 19:39:30 +09:00
  • ea05e3fd5b Merge pull request #1139 from kohya-ss/deep-speed Kohya S 2024-03-26 19:33:57 +09:00
  • a2b8531627 make each script consistent, fix to work w/o DeepSpeed Kohya S 2024-03-25 22:28:46 +09:00
  • c24422fb9d Merge branch 'dev' into deep-speed Kohya S 2024-03-25 22:11:05 +09:00
  • 9c4492b58a fix pytorch version 2.1.1 to 2.1.2 Kohya S 2024-03-24 23:17:25 +09:00
  • 9bbb28c361 update PyTorch version and reorganize dependencies Kohya S 2024-03-24 22:06:37 +09:00
  • 1648ade6da format by black Kohya S 2024-03-24 20:55:48 +09:00
  • 993b2ab4c1 Merge branch 'dev' into deep-speed Kohya S 2024-03-24 18:45:59 +09:00
  • 8d5858826f Merge branch 'dev' into masked-loss Kohya S 2024-03-24 18:19:53 +09:00
  • 025347214d refactor metadata caching for DreamBooth dataset Kohya S 2024-03-24 18:09:32 +09:00
  • ae97c8bfd1 [Experimental] Add cache mechanism for dataset groups to avoid long waiting time for initilization (#1178) Kohaku-Blueleaf 2024-03-24 14:40:18 +08:00
  • 381c44955e update readme and typing hint Kohya S 2024-03-24 11:27:18 +09:00
  • ad97410ba5 Merge pull request #1205 from feffy380/patch-1 Kohya S 2024-03-24 11:14:07 +09:00
  • 691f04322a update readme Kohya S 2024-03-24 11:10:26 +09:00
  • 79d1c12ab0 disable sample_every_n_xxx if value less than 1 ref #1202 Kohya S 2024-03-24 11:06:37 +09:00
  • 0c7baea88c register reg images with correct subset feffy380 2024-03-23 17:28:02 +01:00
  • f4a4c11cd3 support multiline captions ref #1155 Kohya S 2024-03-23 18:51:37 +09:00
  • 594c7f7050 format by black Kohya S 2024-03-23 16:11:31 +09:00
  • d17c0f5084 update dataset config doc Kohya S 2024-03-21 08:31:29 +09:00
  • a35e7bd595 Merge pull request #1200 from BootsofLagrangian/deep-speed Kohya S 2024-03-20 21:32:35 +09:00
  • d9456020d7 Fix most of ZeRO stage uses optimizer partitioning - we have to prepare optimizer and ds_model at the same time. - pull/1139#issuecomment-1986790007 BootsofLagrangian 2024-03-20 20:52:59 +09:00
  • 5095f29e7c fix hf upload #1196 fix-hf-upload Kohya S 2024-03-20 18:46:13 +09:00