Commit Graph

  • fbc8a87a05 Reworked the sd rescaler script Jaret Burkett 2023-08-09 08:57:27 -06:00
  • bf90740b59 Fixed numerous issues with traing ESRGAN Jaret Burkett 2023-08-08 20:03:19 -06:00
  • ff2c9f3d04 Merge branch 'main' of github.com:ostris/ai-toolkit Jaret Burkett 2023-08-07 18:05:03 -06:00
  • 8bd536df7e Added training for a custom version of ERSGAN arcitecture. Testing training now Jaret Burkett 2023-08-07 18:04:23 -06:00
  • 64fbd4c92a Fix some windows dependency issues Jaret Burkett 2023-08-05 20:28:49 -06:00
  • 8c90fa86c6 Complete reqork of how slider training works and optimized it to hell. Can run entire algorythm in 1 batch now with less VRAM consumption than a quarter of it used to take Jaret Burkett 2023-08-05 18:46:08 -06:00
  • 7e4e660663 Added extensions and an example extension that merges models Jaret Burkett 2023-08-04 09:37:24 -06:00
  • b865ac8b24 Various windows bug fixes Jaret Burkett 2023-08-04 05:51:58 -06:00
  • 66c6f0f6f7 Big refactor of SD runner and added image generator Jaret Burkett 2023-08-03 14:51:25 -06:00
  • 75ec5d9292 Hotfix to handle latest transformers clip model missing key suddenly Jaret Burkett 2023-08-02 12:26:18 -06:00
  • 1a25b275c8 Did some work on SD rescaler. Need to run a long test on it eventually. Jaret Burkett 2023-08-02 07:59:27 -06:00
  • 2bf3e529ce Set gradient checkpointing on unet enabled by default. Help out immensly with sdxl backprop spikes Jaret Burkett 2023-08-01 15:43:27 -06:00
  • f53fd08690 Fixed issue with rescaled loras only saving af fp32 Jaret Burkett 2023-08-01 14:08:22 -06:00
  • 8b8d53888d Added Model rescale and prepared a release upgrade WIP Jaret Burkett 2023-08-01 13:49:54 -06:00
  • 63cacf4362 Merge remote-tracking branch 'origin/main' into WIP Jaret Burkett 2023-07-31 15:14:28 -06:00
  • c1b1e800df Updated module names to be compatable with ney koyha unet Jaret Burkett 2023-07-31 13:16:29 -06:00
  • 7726911562 Allow name to be passed through command line Jaret Burkett 2023-07-31 11:56:57 -06:00
  • c01673f1b5 Added random weight adjuster to prevent overfitting Jaret Burkett 2023-07-29 19:30:14 -06:00
  • c35b78f0d4 Added random weight adjuster to prevent overfitting Jaret Burkett 2023-07-29 17:14:14 -06:00
  • 8ba1b11557 Merge branch 'sdxl' into WIP Jaret Burkett 2023-07-29 14:29:18 -06:00
  • 9cdf2dd6e4 Bug fixes sdxl Jaret Burkett 2023-07-29 13:39:57 -06:00
  • 1e50b39442 Work on slider rework Jaret Burkett 2023-07-28 18:11:10 -06:00
  • 5fc2bb5d9c Information trainer Jaret Burkett 2023-07-28 08:16:29 -06:00
  • c7640b0865 WIP diffusers pipeline is weird. Starting to hate sdxl Jaret Burkett 2023-07-27 17:35:24 -06:00
  • b2e2e4bf47 Added sd1.5 and 2.1 do the diffusers pipeline flow Jaret Burkett 2023-07-27 12:34:48 -06:00
  • 596e57a6a6 Pipelines working on SDXL for noise prediction Jaret Burkett 2023-07-27 11:24:33 -06:00
  • 6ab8b8b0f1 WIP. just need to put it here Jaret Burkett 2023-07-27 01:46:30 -06:00
  • 2305e55c82 Moving over to a highly difussers flow for xl Jaret Burkett 2023-07-26 18:25:01 -06:00
  • d3ad195b51 Added rescaling, locon, sdxl, all kinds of stuff. sdxl is still weird Jaret Burkett 2023-07-26 16:19:50 -06:00
  • 40e60fa021 Finially found that bug. ugh Jaret Burkett 2023-07-25 15:28:20 -06:00
  • cb70c03273 SDXL should be working, but I broke something where it is not converging. Jaret Burkett 2023-07-25 13:50:59 -06:00
  • 52f02d53f1 switched to xl branch on sd-scripts Jaret Burkett 2023-07-25 07:41:08 -06:00
  • 7032717294 Added ability to just erase or enhance concepts from a model Jaret Burkett 2023-07-24 17:33:45 -06:00
  • 61dd818608 Added anchors to regulate the lora Jaret Burkett 2023-07-24 14:59:16 -06:00
  • 390192c6a1 Moved the base SD training process to a parent class so we can do some more creative stuff without copy pasta Jaret Burkett 2023-07-24 11:29:39 -06:00
  • e6fb0229bf Added better optimizer chooised and param support Jaret Burkett 2023-07-24 09:21:58 -06:00
  • 9a2819900c added target weight to targets Jaret Burkett 2023-07-23 14:08:37 -06:00
  • 452f2a6da2 Added info, config, etc for lora extracotr and slider trainer Jaret Burkett 2023-07-23 13:13:45 -06:00
  • 9367089d48 Added example for slider training that will run as is Jaret Burkett 2023-07-23 11:24:12 -06:00
  • 434fb22458 Adde dnoise offset Jaret Burkett 2023-07-22 15:01:01 -06:00
  • 3f4f429c4a Reworked the balancing and swapping of the lora during training to make it much more stable when trained Jaret Burkett 2023-07-22 14:13:39 -06:00
  • ce8e7a1271 Added prompt flag to adjust network multiplier Jaret Burkett 2023-07-22 00:05:15 -06:00
  • 596e59dd6d Slider training functioning, time to perfect it Jaret Burkett 2023-07-21 22:06:49 -06:00
  • ddcd9069e1 Base for loopback lora training setup, still working on proper sliders Jaret Burkett 2023-07-21 18:26:02 -06:00
  • aa13251877 Actually addind the Process class..... Jaret Burkett 2023-07-20 15:46:23 -06:00
  • 0761656a90 Added my good ole pattern loss. God I love that thing, conv transpose pattern instantly wiped from vae Jaret Burkett 2023-07-20 15:44:16 -06:00
  • 982e0be7a9 Removed train config, updating it, and added my llvae as pytorch model Jaret Burkett 2023-07-20 14:12:28 -06:00
  • c29b9d075f Setup the base process for merging things. WIP Jaret Burkett 2023-07-20 07:39:31 -06:00
  • 557732e7ff Added Critic support to VAE training. Still tweaking and working on it. Many other fixes Jaret Burkett 2023-07-19 15:57:32 -06:00
  • 6ada328d8d Fixed issues with vgg19 preprocessing. Added yaml support on config file Jaret Burkett 2023-07-19 06:12:57 -06:00
  • 8d8229dfc0 Handle conversions back to ldm for saving Jaret Burkett 2023-07-18 19:34:35 -06:00
  • 17c13eef88 Default to only training mse. Did a lot of cleanup with script. Added logging via tensorboard. Jaret Burkett 2023-07-18 09:40:51 -06:00
  • 94d52572d4 Style and content loss working Jaret Burkett 2023-07-18 07:47:01 -06:00
  • 439310e4dc Made a very basic vae trainer. Jaret Burkett 2023-07-17 19:03:50 -06:00
  • 78b59c5e99 Added support for 3cleir, not fully tested Jaret Burkett 2023-07-16 15:35:14 -06:00
  • 9df7af1720 Adjusted some names and allow to ignore text encoder on extract Jaret Burkett 2023-07-16 12:52:39 -06:00
  • a61914ddd9 Add support for threshold, ratio, and quantile on traditional LoRa Jaret Burkett 2023-07-12 20:02:22 -06:00
  • 8d6edae9fd Added support for traditional LoRa extract using LoCon script Jaret Burkett 2023-07-12 19:51:40 -06:00
  • 57f14e5ef2 WIP implementing training Jaret Burkett 2023-07-12 08:23:46 -06:00
  • 47d094e528 Setup base for training jobs. Added sd-scripts as a submodule Jaret Burkett 2023-07-08 13:50:59 -06:00
  • 37354b006e Reworked so everything is in classes for easy expansion. Single entry point for all config files now. Jaret Burkett 2023-07-08 09:51:42 -06:00
  • 27df03a486 Actually fixed meta data Jaret Burkett 2023-07-05 22:00:58 -06:00
  • 18077ba265 Fixed issue with meta. Jaret Burkett 2023-07-05 16:49:59 -06:00
  • e4de8983c9 Initial commit Jaret Burkett 2023-07-05 16:44:58 -06:00