rockerBOO 6d42b95e2b Refactor transforms, fix loss calculations ...

- add full conditional_loss functionality to wavelet loss
- Transforms are separate and abstracted
- Loss now doesn't include LL except the lowest level
  - ll_level_threshold allows you to control the level the ll is
    used in the loss
- band weights can now be passed in
- rectified flow calculations can be bypassed for experimentation
- Fixed alpha to 1.0 with new weighted bands producing lower loss

2025-05-04 18:39:32 -04:00

..

ipex

Update IPEX libs

2025-02-25 21:27:41 +03:00

__init__.py

Proposed file structure rework and required file changes

2022-12-18 15:24:46 -05:00

adafactor_fused.py

add stochastic rounding, fix single block

2024-08-21 21:04:10 +09:00

attention_processors.py

update diffusers to 1.16 | train_network

2023-06-01 20:39:26 +09:00

config_util.py

Add resize interpolation configuration

2025-02-19 14:20:40 -05:00

custom_offloading_utils.py

feat: add block swap for FLUX.1/SD3 LoRA training

2024-11-12 21:39:13 +09:00

custom_train_functions.py

Refactor transforms, fix loss calculations

2025-05-04 18:39:32 -04:00

deepspeed_utils.py

make deepspeed_utils

2024-02-27 21:30:46 +09:00

device_utils.py

Update IPEX libs

2025-02-25 21:27:41 +03:00

flux_models.py

fix depth value

2024-11-18 13:03:28 +00:00

flux_train_utils.py

Refactor transforms, fix loss calculations

2025-05-04 18:39:32 -04:00

flux_utils.py

add schnell option to load_cn

2024-11-30 00:08:21 +09:00

huggingface_util.py

Replace print with logger if they are logs (#905)

2024-02-04 18:14:34 +09:00

hypernetwork.py

update diffusers to 1.16 | train_network

2023-06-01 20:39:26 +09:00

jpeg_xl_util.py

doc: update license information in jpeg_xl_util.py

2025-03-31 22:02:25 +09:00

lpw_stable_diffusion.py

Replace print with logger if they are logs (#905)

2024-02-04 18:14:34 +09:00

model_util.py

Remove position_ids for V2

2025-02-20 04:49:51 -05:00

original_unet.py

Replace print with logger if they are logs (#905)

2024-02-04 18:14:34 +09:00

sai_model_spec.py

refactor SD3 CLIP to transformers etc.

2024-10-24 19:49:28 +09:00

sd3_models.py

fix: resolve model corruption issue with pos_embed when using --enable_scaled_pos_embed

2024-12-07 15:12:27 +09:00

sd3_train_utils.py

feat: add block swap for FLUX.1/SD3 LoRA training

2024-11-12 21:39:13 +09:00

sd3_utils.py

support SD3.5M

2024-10-30 12:51:49 +09:00

sdxl_lpw_stable_diffusion.py

sample generation in SDXL ControlNet training

2024-09-30 23:39:32 +09:00

sdxl_model_util.py

use original ControlNet instead of Diffusers

2024-09-29 23:07:34 +09:00

sdxl_original_control_net.py

use original ControlNet instead of Diffusers

2024-09-29 23:07:34 +09:00

sdxl_original_unet.py

use original ControlNet instead of Diffusers

2024-09-29 23:07:34 +09:00

sdxl_train_util.py

Merge branch 'dev' into sd3

2025-02-24 18:56:15 +09:00

slicing_vae.py

Replace print with logger if they are logs (#905)

2024-02-04 18:14:34 +09:00

strategy_base.py

Support SD3.5M multi resolutional training

2024-10-31 19:58:22 +09:00

strategy_flux.py

Support SD3.5M multi resolutional training

2024-10-31 19:58:22 +09:00

strategy_sd3.py

Support SD3.5M multi resolutional training

2024-10-31 19:58:22 +09:00

strategy_sd.py

Fix training, validation split, revert to using upstream implemenation

2025-01-03 15:20:25 -05:00

strategy_sdxl.py

support weighted captions for SD/SDXL

2024-10-11 08:48:55 +09:00

train_util.py

Add guidance scale for prompt param and flux sampling

2025-04-16 21:50:36 +09:00

utils.py

Fix the interpolation logic error in resize_image()

2025-04-02 22:04:37 +08:00