Lincoln Stein
aa247e68be
use refined model by default
2022-11-02 18:29:34 -04:00
Lincoln Stein
895c47fd11
Merge branch 'patch-9' of https://github.com/spezialspezial/stable-diffusion into spezialspezial-patch-9
2022-11-02 18:24:55 -04:00
damian
d85cd99f17
add option to show intermediate latent space
2022-11-02 17:53:11 -04:00
spezialspezial
825fa6977d
Update outcrop.py
2022-11-02 16:33:35 -04:00
spezialspezial
e332529fbd
Prevent outcrop error when no callback is supplied
2022-11-02 16:33:35 -04:00
damian0815
2468a28e66
save VRAM by not recombining tensors that have been sliced to save VRAM
2022-11-01 22:39:48 -04:00
damian0815
e3ed748191
fix a bug that broke cross attention control index mapping
2022-11-01 22:39:39 -04:00
damian0815
3f5bf7ac44
report full size for fast latents and update conversion matrix for v1.5
2022-11-01 22:39:27 -04:00
spezialspezial
b45e632f23
Option to directly invert the grayscale heatmap - fix
2022-11-01 22:18:00 -04:00
spezialspezial
6c9a2761f5
Optional refined model for Txt2Mask
...
Don't merge right now, just wanted to show the necessary changes
2022-11-02 00:33:46 +01:00
spezialspezial
2bdd738f03
Update txt2mask.py
2022-11-01 17:39:56 -04:00
spezialspezial
7782760541
Option to directly invert the grayscale heatmap
...
Theoretically less work inverting the image while it's small but I can't measure a significant difference. Though, handy option to have in some cases.
2022-11-01 17:39:56 -04:00
damian0815
de2686d323
fix crash (be a little less aggressive clearing out the attention slice)
2022-11-01 17:35:43 -04:00
damian0815
0b72a4a35e
be more aggressive at clearing out saved_attn_slice
2022-11-01 17:35:34 -04:00
Lincoln Stein
942a202945
fix model_cache memory management issues
2022-11-01 17:22:48 -04:00
Lincoln Stein
89da42ad79
Merge branch 'pin-options-panel' of https://github.com/psychedelicious/stable-diffusion into psychedelicious-pin-options-panel
...
- from PR #1301
2022-10-31 09:37:13 -04:00
Damian at mba
ced9c83e96
various prompting fixes
2022-10-31 09:34:56 -04:00
Lincoln Stein
80f2cfe3e3
set default max_models to 2 internally as well as as arg
2022-10-31 09:05:38 -04:00
Lincoln Stein
dc556cb1a7
add max_load_models parameter for model cache control
...
- ldm.generate.Generator() now takes an argument named `max_load_models`.
This is an integer that limits the model cache size. When the cache
reaches the limit, it will start purging older models from cache.
- CLI takes an argument --max_load_models, default to 2. This will keep
one model in GPU and the other in CPU and switch back and forth
quickly.
- To not cache models at all, pass --max_load_models=1
2022-10-31 08:55:53 -04:00
Lincoln Stein
0c8f0e3386
add max_load_models parameter for model cache control
...
- ldm.generate.Generator() now takes an argument named `max_load_models`.
This is an integer that limits the model cache size. When the cache
reaches the limit, it will start purging older models from cache.
- CLI takes an argument --max_load_models, default to 2. This will keep
one model in GPU and the other in CPU and switch back and forth
quickly.
- To not cache models at all, pass --max_load_models=1
2022-10-31 08:53:16 -04:00
Lincoln Stein
98f03053ba
hard-code strength to 0.9 during outcropping
2022-10-31 07:52:34 -04:00
Lincoln Stein
59ef2471e1
improve outcropping performance
...
- applied inpainting parameters recommended by @kyle0654
- results are aesthetically pleasing
- Closes #1319
2022-10-31 07:52:26 -04:00
Lincoln Stein
ce7651944d
adapt outcrop.py to use new outpainting code
...
- unfortunately it does not look as good as the old code
which just used plain inpainting.
2022-10-31 07:52:13 -04:00
Lincoln Stein
a3e0b285d8
fix embiggen crash
2022-10-31 07:52:06 -04:00
Lincoln Stein
3cdfedc649
hard-code strength to 0.9 during outcropping
2022-10-31 01:54:32 -04:00
Lincoln Stein
531f596bd1
improve outcropping performance
...
- applied inpainting parameters recommended by @kyle0654
- results are aesthetically pleasing
- Closes #1319
2022-10-31 01:37:12 -04:00
Lincoln Stein
8683426041
add seamless to metadata
2022-10-31 00:40:30 -04:00
Lincoln Stein
582fee6c3a
adapt outcrop.py to use new outpainting code
...
- unfortunately it does not look as good as the old code
which just used plain inpainting.
2022-10-31 00:20:53 -04:00
Lincoln Stein
2b39d1677c
fix embiggen crash
2022-10-30 22:57:15 -04:00
Lincoln Stein
23d54ee69e
fix mps crash with safety checker
2022-10-30 16:54:06 -04:00
Lincoln Stein
f70af7afb9
remove debug image gen from outcrop
2022-10-30 12:19:43 -04:00
Lincoln Stein
e7368d7231
preload_models interactively downloads sd model files
2022-10-30 12:19:05 -04:00
Lincoln Stein
231dfe01f4
fix incorrect thresholding reporting for karras noise; close #1300
2022-10-30 10:35:55 -04:00
spezialspezial
c2fab45a6e
Prevent indexing error for mode RGB
...
I have not explicitly tested mode P
2022-10-29 18:20:53 -04:00
Lincoln Stein
0f4413da7d
Merge branch 'inpainting-rebase' of https://github.com/psychedelicious/stable-diffusion into psychedelicious-inpainting-rebase
2022-10-29 13:42:00 -04:00
Lincoln Stein
fdf9b1c40c
fix CLI inpainting crash
2022-10-29 11:47:06 -04:00
Lincoln Stein
13f26a99b8
documentation and usability fixes
2022-10-29 10:37:38 -04:00
Lincoln Stein
ef68a419f1
preload_models.py script downloads the weight files
...
- user can select which weight files to download using huggingface cache
- user must log in to huggingface, generate an access token, and accept
license terms the very first time this is run. After that, everything
works automatically.
- added placeholder for docs for installing models
- also got rid of unused config files. hopefully they weren't needed
for textual inversion, but I don't think so.
2022-10-29 01:02:45 -04:00
Kyle Schouviller
43de16cae4
Don't try to tile fill if image doesn't have an alpha layer
2022-10-29 04:25:27 +11:00
Damian at mba
864d080502
handle all unicode characters
2022-10-28 10:39:12 -04:00
Lincoln Stein
05b8de5300
fix --hires to support inpainting model
2022-10-27 23:12:21 -04:00
Lincoln Stein
387f796ebe
Merge branch 'development' into development
2022-10-27 23:04:04 -04:00
Lincoln Stein
3033331f65
remove unneeded warnings from attention.py
2022-10-27 22:50:06 -04:00
Lincoln Stein
362b234cd1
fix long-standing issue with metadata retrieval
...
The Args object would crap out when trying to retrieve metadata from
an image file that did not contain InvokeAI-generated metadata, such
as a JPG. This corrects that and returns dummy values (seed of zero,
prompt of '') to avoid downstream breakage.
2022-10-27 22:43:34 -04:00
Lincoln Stein
2115874587
resolve conflicts with outpainting implementation
2022-10-27 18:06:38 -04:00
Lincoln Stein
cd5141f3d1
fix issues with outpaint merge
2022-10-27 18:02:08 -04:00
Lincoln Stein
b815aa2130
Merge branch 'development' into outpaint
2022-10-27 17:17:34 -04:00
Lincoln Stein
19a6e904ec
resolved whitespace difference
2022-10-27 17:12:22 -04:00
Lincoln Stein
1200fbd3bd
add threshold for switchover from Karras to LDM noise schedule
2022-10-27 17:07:50 -04:00
Lincoln Stein
d25bf7a55a
cut over from karras to model noise schedule for higher steps
...
The k_samplers come with a "karras" noise schedule which performs
very well at low step counts but becomes noisy at higher ones.
This commit introduces a threshold (currently 30 steps) at which the
k samplers will switch over from using karras to the older model
noise schedule.
2022-10-27 17:06:49 -04:00
