Commit Graph

2184 Commits

Author SHA1 Message Date
rockerBOO
d40f5b1e4e Revert "Scale sigmoid to default 1.0"
This reverts commit 8aa126582e.
2025-03-20 15:09:50 -04:00
rockerBOO
8aa126582e Scale sigmoid to default 1.0 2025-03-20 15:09:11 -04:00
rockerBOO
e8b3254858 Add flux_train_utils tests for get get_noisy_model_input_and_timesteps 2025-03-20 15:01:15 -04:00
rockerBOO
16cef81aea Refactor sigmas and timesteps 2025-03-20 14:32:56 -04:00
Kohya S
d151833526 docs: update README with recent changes and specify version for pytorch-optimizer 2025-03-20 22:05:29 +09:00
Kohya S.
936d333ff4 Merge pull request #1985 from gesen2egee/pytorch-optimizer
Support pytorch_optimizer
2025-03-20 22:01:03 +09:00
rockerBOO
f974c6b257 change order to match upstream 2025-03-19 14:27:43 -04:00
rockerBOO
5d5a7d2acf Fix IP noise calculation 2025-03-19 13:50:04 -04:00
rockerBOO
1eddac26b0 Separate random to a variable, and make sure on device 2025-03-19 00:49:42 -04:00
rockerBOO
8e6817b0c2 Remove double noise 2025-03-19 00:45:13 -04:00
rockerBOO
d93ad90a71 Add perturbation on noisy_model_input if needed 2025-03-19 00:37:27 -04:00
rockerBOO
7197266703 Perturbed noise should be separate of input noise 2025-03-19 00:25:51 -04:00
gesen2egee
5b210ad717 update prodigyopt and prodigy-plus-schedule-free 2025-03-19 10:49:06 +08:00
rockerBOO
b81bcd0b01 Move IP noise gamma to noise creation to remove complexity and align noise for target loss 2025-03-18 21:36:55 -04:00
rockerBOO
6f4d365775 zeros_like because we are adding 2025-03-18 18:53:34 -04:00
rockerBOO
a4f3a9fc1a Use ones_like 2025-03-18 18:44:21 -04:00
rockerBOO
b425466e7b Fix IP noise gamma to use random values 2025-03-18 18:42:35 -04:00
rockerBOO
c8be141ae0 Apply IP gamma to noise fix 2025-03-18 15:42:18 -04:00
rockerBOO
0b25a05e3c Add IP noise gamma for Flux 2025-03-18 15:40:40 -04:00
rockerBOO
3647d065b5 Cache weight norms estimate on initialization. Move to update norms every step 2025-03-18 14:25:09 -04:00
Disty0
620a06f517 Check for uppercase file extension too 2025-03-17 17:44:29 +03:00
Disty0
564ec5fb7f use extend instead of += 2025-03-17 17:41:03 +03:00
Disty0
7e90cdd47a use bytearray and add typing hints 2025-03-17 17:26:08 +03:00
gesen2egee
e5b5c7e1db Update requirements.txt 2025-03-15 13:29:32 +08:00
rockerBOO
ea53290f62 Add LoRA-GGPO for Flux 2025-03-06 00:00:38 -05:00
Kohya S.
75933d70a1 Merge pull request #1960 from kohya-ss/sd3_safetensors_merge
Sd3 safetensors merge
2025-03-05 23:28:38 +09:00
Kohya S
aa2bde7ece docs: add utility script for merging SD3 weights into a single .safetensors file 2025-03-05 23:24:52 +09:00
sdbds
3f49053c90 fatser fix bug for SDXL super SD1.5 assert cant use 32 2025-03-02 19:32:06 +08:00
Ivan Chikish
acdca2abb7 Fix [occasionally] missing text encoder attn modules
Should fix #1952
I added alternative name for CLIPAttention.
I have no idea why this name changed.
Now it should accept both names.
2025-03-01 20:35:45 +03:00
Kohya S
ba5251168a fix: save tensors as is dtype, add save_precision option 2025-03-01 10:31:39 +09:00
Kohya S
272f4c3775 Merge branch 'sd3' into sd3_safetensors_merge 2025-02-28 23:52:36 +09:00
Kohya S
734333d0c9 feat: enhance merging logic for safetensors models to handle key prefixes correctly 2025-02-28 23:52:29 +09:00
Disty0
2f69f4dbdb fix typo 2025-02-27 00:30:19 +03:00
Disty0
9a415ba965 JPEG XL support 2025-02-27 00:21:57 +03:00
Kohya S
3d79239be4 docs: update README to include recent improvements in validation loss calculation 2025-02-26 21:21:04 +09:00
Kohya S
ec350c83eb Merge branch 'dev' into sd3 2025-02-26 21:17:29 +09:00
Kohya S.
49651892ce Merge pull request #1903 from kohya-ss/val-loss-improvement
Val loss improvement
2025-02-26 21:15:14 +09:00
Kohya S
1fcac98280 Merge branch 'sd3' into val-loss-improvement 2025-02-26 21:09:10 +09:00
Kohya S.
b286304e5f Merge pull request #1953 from Disty0/dev
Update IPEX libs
2025-02-26 21:03:09 +09:00
Kohya S
ae409e83c9 fix: FLUX/SD3 network training not working without caching latents closes #1954 2025-02-26 20:56:32 +09:00
Kohya S
5228db1548 feat: add script to merge multiple safetensors files into a single file for SD3 2025-02-26 20:50:58 +09:00
Kohya S
f4a0047865 feat: support metadata loading in MemoryEfficientSafeOpen 2025-02-26 20:50:44 +09:00
Disty0
f68702f71c Update IPEX libs 2025-02-25 21:27:41 +03:00
Kohya S.
6e90c0f86c Merge pull request #1909 from rockerBOO/progress_bar
Move progress bar to account for sampling image first
2025-02-24 18:57:44 +09:00
Kohya S
67fde015f7 Merge branch 'dev' into sd3 2025-02-24 18:56:15 +09:00
Kohya S.
386b7332c6 Merge pull request #1918 from tsukimiya/fix_vperd_warning
Remove v-pred warning.
2025-02-24 18:55:25 +09:00
Kohya S
905f081798 Merge branch 'dev' into sd3 2025-02-24 18:54:28 +09:00
Kohya S.
59ae9ea20c Merge pull request #1945 from yidiq7/dev
Remove position_ids for V2
2025-02-24 18:53:46 +09:00
Kohya S
efb2a128cd fix wandb val logging 2025-02-21 22:07:35 +09:00
Yidi
13df47516d Remove position_ids for V2
The postions_ids cause errors for the newer version of transformer.
This has already been fixed in convert_ldm_clip_checkpoint_v1() but
not in v2.
The new code applies the same fix to convert_ldm_clip_checkpoint_v2().
2025-02-20 04:49:51 -05:00