a8fba9af35
medvram support for SD3
2024-06-24 10:15:46 +03:00
5b2a60b8e2
initial SD3 support
2024-06-16 08:04:31 +03:00
45b8a499a7
fix wrong condition
2024-03-02 10:36:48 +03:00
bb24c13ed7
infotext support for #14978
2024-03-02 07:39:59 +03:00
bf348032bc
fix missing arg
2024-02-20 16:59:28 -05:00
25eeeaa65f
Allow refiner to be triggered by model timestep instead of sampling
2024-02-20 16:37:29 -05:00
757dda9ade
Add Pad conds v0 option
2024-01-27 22:30:47 +03:00
d43333ff71
fix an issue where VAE would remain in fp16 after an auto-switch to fp32
2023-08-30 21:13:24 +03:00
87cca029d7
add an option to choose how to combine hires fix and refiner
2023-08-30 18:24:21 +03:00
a459075d26
actual solution to the uncommon hanging problem that is seemingly caused by multiple progress requests working on same tensor
2023-08-22 10:41:10 +03:00
d7c9c61420
attemped solution to the uncommon hanging problem that is seemingly caused by live previews working on the tensor as denoising
2023-08-22 09:55:20 +03:00
953c3eab7b
forbid Full live preview method for medvram and add a setting to undo the forbidding
2023-08-21 15:54:30 +03:00
7056fdf2be
Merge pull request #12630 from catboxanon/fix/nans-mk2
...
Attempt to resolve NaN issue with unstable VAEs in fp32 mk2
2023-08-19 08:34:46 +03:00
3ce5fb8e5c
Add option for faster live interrupt
2023-08-17 20:03:26 -04:00
3003b10e0a
Attempt to resolve NaN issue with unstable VAEs in fp32 mk2
2023-08-17 18:10:55 -04:00
d9ddc5d4cd
Remove wrong scale
2023-08-16 11:21:12 +08:00
45be87afc6
correctly add Eta DDIM to infotext when it's 1.0 and do not add it when it's 0.0.
2023-08-14 21:48:05 +03:00
127ab9114f
Merge pull request #12514 from catboxanon/feat/batch-encode
...
Encode batch items individually to significantly reduce VRAM
2023-08-13 16:41:07 +03:00
d1a70c3f05
Add s_noise param to more samplers
2023-08-13 08:22:24 -04:00
822597db49
Encode batches separately
...
Significantly reduces VRAM.
This makes encoding more inline with how decoding currently functions.
2023-08-13 04:16:48 -04:00
599f61a1e0
use dataclass for StableDiffusionProcessing
2023-08-13 08:24:16 +03:00
fa9370b741
add refiner to StableDiffusionProcessing class
...
write out correct model name in infotext, rather than the refiner model
2023-08-13 06:07:30 +03:00
b293ed3061
make it possible to use hires fix together with refiner
2023-08-12 12:54:32 +03:00
64311faa68
put refiner into main UI, into the new accordions section
...
add VAE from main model into infotext, not from refiner model
option to make scripts UI without gr.Group
fix inconsistencies with refiner when usings samplers that do more denoising than steps
2023-08-12 12:39:59 +03:00
70a01cd444
Merge branch 'dev' into refiner
2023-08-10 17:04:38 +03:00
0d5dc9a6e7
rework RNG to use generators instead of generating noises beforehand
2023-08-09 08:43:31 +03:00
1aefb50259
add None refiner option
2023-08-08 22:17:25 +03:00
f8ff8c0638
merge errors
2023-08-08 22:09:51 +03:00
54c3e5c913
Merge branch 'dev' into refiner
2023-08-08 21:49:47 +03:00
8285a149d8
add CFG denoiser implementation for DDIM, PLMS and UniPC (this is the commit when you can run both old and new implementations to compare them)
2023-08-08 21:04:44 +03:00
5a0db84b6c
add infotext
...
add proper support for recalculating conds in k-diffusion samplers
remove support for compvis samplers
2023-08-06 17:53:33 +03:00
f1975b0213
initial refiner support
2023-08-06 17:01:07 +03:00
aa42c0ff8e
repair broken live previews if using VAE with half
2023-08-06 07:41:24 +03:00
073342c887
remove noneed scale
2023-08-04 17:55:52 +08:00
6346d8eeaa
Revert "change all encode"
...
This reverts commit 094c416a80
2023-08-04 17:53:30 +08:00
094c416a80
change all encode
2023-08-04 17:53:16 +08:00
1f6bfdea80
move the modified decode into smapler_common
2023-08-04 14:38:52 +08:00
70e66e81e5
Merge branch 'dev' into efficient-vae-methods
2023-08-04 14:38:16 +08:00
f0c1063a70
resolve some of circular import issues for kohaku
2023-08-04 09:13:46 +03:00
c134a48016
Fix code style
2023-08-04 13:40:20 +08:00
75336dfc84
add TAESD for i2i and t2i
2023-08-04 13:38:52 +08:00
fca42949a3
rework torchsde._brownian.brownian_interval replacement to use device.randn_local and respect the NV setting.
2023-08-03 07:18:55 +03:00
1210548cba
simplify single_sample_to_image
2023-05-17 14:53:39 +03:00
7a13a3f4ba
TAESD fix
2023-05-17 17:39:07 +08:00
85232a5b26
Merge branch 'dev' into taesd-a
2023-05-17 09:26:26 +03:00
56a2672831
return live preview defaults to how they were
...
only download TAESD model when it's needed
return calculations in single_sample_to_image to just if/elif/elif blocks
keep taesd model in its own directory
2023-05-17 09:24:01 +03:00
cdac5ace14
suppress ENSD infotext for samplers that don't use it
2023-05-16 11:54:02 +03:00
bd9b9d425a
Add live preview mode check
2023-05-14 14:06:02 +08:00
e14b586d04
Add Tiny AE live preview
2023-05-14 14:06:01 +08:00
5fe0dd79be
rename CPU RNG to RNG source in settings, add infotext and parameters copypaste support to RNG source
2023-04-29 11:29:37 +03:00
AUTOMATIC1111