ZeroCool940711
d8a8687852
Improved img2txt layout and performance.
2022-09-30 12:41:01 -07:00
ZeroCool940711
3bb432162e
Added open clip dependency which is needed for some CLIP models.
2022-09-30 08:47:30 -07:00
hlky
084fa9732f
Update img2txt.py
2022-09-30 10:40:02 +01:00
ZeroCool940711
51b7c9ac31
Removed some left over comments on webui_streamlit.py
2022-09-30 02:05:16 -07:00
ZeroCool940711
8673363d71
Moved the css loading to be on sd_utils.py
2022-09-30 02:04:52 -07:00
ZeroCool940711
50bc2bb545
Fixed typo on the model manager
2022-09-30 02:04:32 -07:00
ZeroCool940711
5445287bcf
Img2txt working now but needs more than 8GB og VRAM to work. Will be trying to improve it as the next step.
2022-09-29 08:52:46 -07:00
ZeroCool940711
466cfde3fa
More fixes for img2txt
2022-09-29 07:27:56 -07:00
ZeroCool940711
625ba555e3
Fixed custom models not loading on txt2img, img2img and txt2vid.
2022-09-29 03:29:44 -07:00
ZeroCool940711
ae263cb55e
Made sure the txt2vid tab shows a loading animation when the models are loading.
2022-09-29 01:07:32 -07:00
ZeroCool940711
07d935b445
More fixes to img2txt
2022-09-29 01:07:32 -07:00
Alejandro Gil
f678efb4e2
Img2txt dependencies and necessary files. ( #1354 )
2022-09-28 12:37:15 -07:00
Alejandro Gil
510d103a83
Temporary fix for the jumpy UI and tabs resetting to the first one. ( #1353 )
2022-09-28 10:02:08 -07:00
Alejandro Gil
3e40bab7f7
Changed the default streamlit import for hydralit as we will be using hydralit as replacement for the default streamlit library, hydralit provides better control over css as well as having a lot more options. ( #1352 )
2022-09-28 09:33:54 -07:00
Alejandro Gil
1f4b564de4
Removed the models in case of OOM so we can reload them. ( #1343 )
2022-09-27 11:51:17 -07:00
Alejandro Gil
519d7c661f
Added new implementation of the textual inversion page. ( #1342 )
2022-09-27 11:36:48 -07:00
ZeroCool
8f2ed2f350
Added message on the text to video tab for specifying how to add the huggingface token if its missing. ( #1339 )
...
# Checklist:
- [x] I have changed the base branch to `dev`
- [x] I have performed a self-review of my own code
- [x] I have commented my code in hard-to-understand areas
- [x] I have made corresponding changes to the documentation
2022-09-27 00:57:28 -07:00
Lizardy
62790406bf
Fix --config args
...
* Fixed config args being double-loaded and not replacing the defaults.
2022-09-26 23:58:41 -04:00
ZeroCool
6db88c4673
Merge branch 'dev' into dev
2022-09-26 07:03:59 -07:00
ZeroCool
32b8714b1d
Delete daisi_app.py
2022-09-26 07:03:33 -07:00
ZeroCool940711
7bd036a279
Added config.toml file again.
2022-09-26 06:53:14 -07:00
ZeroCool940711
b06199ebb6
Added expander for the variant option on txt2vid to reduce the amount of space it uses.
2022-09-26 06:17:50 -07:00
ZeroCool940711
35ec438682
Daisi app fix 4
...
Added option on the settings page to add your huggingface token.
Daisi app fix 3
Daisi app fix 2
Daisi app fix
2022-09-26 06:08:13 -07:00
hlky
4a15702fe9
License header
2022-09-26 14:02:48 +01:00
ZeroCool940711
61e4839a1b
Merge remote-tracking branch 'origin/dev' into dev
2022-09-26 05:00:34 -07:00
ZeroCool940711
d940c491a7
- Improved the layout of the UI.
...
- Moved relevant options to the top and group them together for easy access.
2022-09-26 04:53:34 -07:00
ZeroCool940711
e5e4954aeb
Implemented daisi app base loader for the streamlit UI.
2022-09-26 04:39:40 -07:00
ZeroCool940711
2c8ea3d103
Updated the links on the model manager to comply with HuggingFace's TOS.
2022-09-26 04:17:06 -07:00
ZeroCool940711
0d9f3f4947
Added support for running the webui on daisi.io
2022-09-26 02:36:21 -07:00
ZeroCool940711
0b6d0d3b23
The default streamlit theme can now be customized from the UI as well as the UI config file.
2022-09-26 01:24:14 -07:00
ZeroCool940711
3b4b435768
Added option to enable or disable the streamlit data usage gathering or telemetry.
2022-09-26 01:24:14 -07:00
ZeroCool940711
eb9145000a
The default streamlit theme can now be customized from the UI as well as the UI config file.
2022-09-26 01:20:31 -07:00
ZeroCool940711
c37a92c39f
Added option to enable or disable the streamlit data usage gathering or telemetry.
2022-09-26 01:12:06 -07:00
ZeroCool
1fd28eed1e
Made sure GFPGAN and RealESRGAN are on server_state. ( #1319 )
2022-09-25 21:23:32 -07:00
ZeroCool940711
e6529b22ec
Made sure GFPGAN and RealESRGAN are on server_state.
2022-09-25 21:07:06 -07:00
Qualzz
5d1558f2b4
Concept library improvements ( #1314 )
...
* corrupted pngs handling
* hide search label
* clear cache button
2022-09-25 20:23:19 +01:00
Chris Heald
c7e126f6f8
Updates for the v2 trainer
...
* Restore --config. This will be useful when you have an init config
that you don't want overwritten.
* Cache the individual transformed images in TextualInversionDataset.
This gains speed by avoiding reading and reprocessing the image each
time it's used for training.
* Turn on no_grad for inference and clean up tensors during
checkpointing. This reduces memory usage slightly.
* Set the sample output size to 384x384. We just need them large enough
for manual evaluation, and this gains us a decent chunk of speed.
* (breaking change) Custom templates are now semicolon-delineated.
Additionally, custom templates are properly passed through to
TextualInversionDataset to generate input_ids for your images. Using
custom templates which accurately describe your input images seems to
improve training fidelity.
* Cache autoencoding of image pixel data. This substantially increases
the speed of training, upwards of 40% for me.
* Clean up a little bit of cruft.
2022-09-25 13:21:24 +01:00
hlky
0d79243e45
fix diffusers_textual_inversion_2.py conflicts step 1
2022-09-25 13:20:17 +01:00
ZeroCool940711
a1fff59a73
Fixed "device" not being on server_state.
2022-09-25 03:55:16 -07:00
ZeroCool940711
47df4681ee
Reduced the folder name a bit more to help with long paths on windows.
2022-09-25 01:44:34 -07:00
ZeroCool940711
4e34a987e1
Added customizable value for the concepts per page option on the concepts library tab.
2022-09-25 01:17:14 -07:00
ZeroCool940711
7084e94bdf
Moved the model to server_state instead of session_state so it can be shared between multiple sessions, tabs and users as long as the streamlit server is running.
2022-09-25 00:03:05 -07:00
ZeroCool940711
86505f16ed
Moved the pipe model to server_state instead of session_state so it can be shared between multiple sessions, tabs and users as long as the streamlit server is running.
2022-09-24 23:28:02 -07:00
ZeroCool940711
7c24e6f620
Moved the models as well as other static parameters to the server_state instead of session_state.
2022-09-24 23:23:55 -07:00
ZeroCool
a64f9a22b0
Merge branch 'dev' into img2img-esrgan-fix
2022-09-24 18:26:54 -07:00
ZeroCool940711
d7d0ad97c3
Changed some PIL stuff to use resampling to avoid some warnings on the console
2022-09-24 18:06:24 -07:00
ZeroCool940711
9c12a6a322
Changed the search bar text_input widget to have a label as its not allowed to be empty on streamlit 1.13.0 and will give a warning and possible an error later.
2022-09-24 18:05:08 -07:00
ZeroCool940711
b8b9df717a
Removed some unused functions and code so we can upgrade streamlit to version 1.13.0.
2022-09-24 18:04:21 -07:00
Seacant
bb029225d6
Fixed Streamlit bug where RealESRGAN seemed broke
...
There was a safety check where RealESRGAN and loopback
were not allowed to be on at the same time to prevent
rapidly growing images (I think, at least) which was
causing confusion in the UI since there was no indication
this wasn't allowed. Using ESRGAN just on the final iteration
should be safe, so this commit enables that.
2022-09-25 01:01:25 +00:00
ZeroCool940711
23070de38d
Added info text to let the user now that the other setting pages are still under construction.
2022-09-24 17:18:29 -07:00
ZeroCool940711
3816b28d2c
Fixed error with CustomModel_available not being in session_state after the Settings page change.
2022-09-24 15:27:09 -07:00
ZeroCool940711
0f9f7c382b
First implementation of the Settings page.
2022-09-24 05:30:16 -07:00
ZeroCool940711
37b2090844
Added message to tell user that the textual inversion page is under construction instead of having it empty.
2022-09-24 05:28:54 -07:00
ZeroCool940711
d584fa28eb
Added condition to make sure the userconfig_streamlit.yaml file is created if it doesnt exist.
2022-09-24 05:19:13 -07:00
ZeroCool940711
6a7de2066f
Removed script from before the code split that is no longer needed.
2022-09-24 05:15:18 -07:00
ZeroCool940711
9b73f463cf
- Added function to convert bytes to a human readable string.
...
- Added basic implementation of "enable_minimal_memory_usage".
2022-09-24 05:11:33 -07:00
ZeroCool940711
1c459bce53
- Added individual titles on the browser tab for each page.
...
- Added basic url for each page to the address bar for browser history.
2022-09-23 22:07:50 -07:00
ZeroCool940711
8cf8d3f03a
Added simple function to allows to change the title of the app dynamically instead of only doing it once per app, this could be useful for setting the title on specific pages and making it appear on the browser history.
2022-09-23 21:19:50 -07:00
xaedes
70acb61c21
remove wrong arguments from save_sample call ( #1290 )
...
fixes #1258
2022-09-24 01:29:12 +01:00
Qualzz
6ae84c6c3a
better library browser ( #1281 )
...
* better library browser
* removed debug prints
2022-09-24 00:20:51 +01:00
hlky
f6215a5bb1
remove unused scripts
2022-09-23 15:31:07 +01:00
hlky
8b4464a706
Add periodic checkpointing
...
Update diffusers_textual_inversion_2.py
2022-09-23 15:29:08 +01:00
Colton
f0be43d7df
Send the image grid to the front end if generated for a prompt matrix ( #1200 )
2022-09-23 14:32:36 +03:00
Qualzz
75eac9c5c0
moved pagination logic before the rendering of the component ( #1268 )
2022-09-22 23:09:42 +03:00
ZeroCool940711
e49699a65e
Basic code for the streamlit textual inversion page.
2022-09-22 01:44:35 -07:00
ZeroCool940711
889256d26e
Improved the script that downloads the concepts library from huggingface in case someone wants to use it.
2022-09-22 00:02:44 -07:00
ZeroCool940711
3cd91bde82
The denoising_strength slider on img2img is now customizable using the config file.
2022-09-21 14:58:49 -07:00
ZeroCool940711
2c7b22a540
The Find Noise Steps slider om img2img can now be customizable from the config file.
2022-09-21 00:39:27 -07:00
ZeroCool940711
e874a2d6a7
Added condition to check if the concepts library folder exist and if not tell the user where to download the concepts from.
2022-09-20 22:15:49 -07:00
ZeroCool940711
8584205e11
Added implementation of the Concepts Library tab made by @Qualzz.
2022-09-20 21:06:00 -07:00
ZeroCool940711
f99657241d
Fixed the image not been decoded in txt2vid if the Update Image Preview options was not checked.
2022-09-20 17:07:38 -07:00
hlky
ba84d28da2
txt2vid fix enumerate(... max_frames)) -> enumerate(... num_steps))
2022-09-20 14:00:08 +01:00
Thomas Mello
3705f296c6
fix: division by zero, inefficient preview check ( #1230 )
2022-09-19 21:05:18 +03:00
hlky
c6bf0d2ecb
Update diffusers_textual_inversion_2.py
2022-09-19 14:49:00 +01:00
hlky
91c0d0f20b
Create diffusers_textual_inversion_2.py
2022-09-19 13:09:45 +01:00
ZeroCool940711
56ae22eb53
- Improved optimized mode for txt2img, it should be able to generate high resolution images using less VRAM than before.
...
- Improved txt2vid so its now possible to generate high resolution images with less VRAM than before.
- Added condition to ensure that the pipe model from txt2vid is removed from memory when switching from the txt2vid tab to txt2img and viceversa.
2022-09-19 00:02:21 -07:00
ZeroCool940711
1d9ef57802
Fixed the filename of the video generated at the end of a txt2vid generation, it was saved as test.mp before but now its saved as seed_prompt.mp4 with the seed and prompt form the generation.
2022-09-18 18:29:19 -07:00
ZeroCool940711
a7d7955721
GFPGAN can now be used on the txt2vid tab.
2022-09-18 15:17:02 -07:00
ZeroCool940711
db0eb80728
Fixed typo on the sd-concepts-library path.
2022-09-18 14:20:13 -07:00
ZeroCool940711
1b1f784a76
Every slider min, max and step values as well as its default value is now configurable through the config file.
2022-09-18 11:11:23 -07:00
ZeroCool940711
fdd0c38ac2
Added missing sd_concept_library.py file.
2022-09-18 09:15:05 -07:00
hlky
6f4a1d8a41
slider_steps and slider_bounds in defaults config
...
slider_steps and slider_bounds in defaults config
2022-09-18 12:23:51 +01:00
Thomas Mello
afe954d4e9
Merge remote-tracking branch 'upstream/dev' into the-merge
2022-09-18 13:03:19 +03:00
Dekker3D
56b91b8bdc
Streamlit txt2img page settings now follow defaults ( #1195 )
...
* Some options on the Streamlit txt2img page now follow the defaults from the relevant config files.
* Fixed a copy-paste gone wrong in my previous commit.
* st.session_state["defaults"] fix
Co-authored-by: hlky <106811348+hlky@users.noreply.github.com>
2022-09-18 10:39:30 +01:00
protoplm
e4d82f4e84
Used loaded_model state variable in .yaml generation ( #1196 )
...
Used loaded_model state variable in .yaml generation
2022-09-18 10:34:37 +01:00
Thomas Mello
148c41826b
Merge branch 'dev' into the-merge
...
# Conflicts:
# .gitignore
# README.md
# configs/webui/webui_streamlit.yaml
# environment.yaml
# frontend/css/streamlit.main.css
# frontend/frontend.py
# scripts/webui_streamlit.py
2022-09-18 12:31:17 +03:00
hlky
4680975ae1
st.session_state["defaults"] fix
2022-09-18 09:57:33 +01:00
hlky
745e1cc2e1
Temporarily hide sd_concept_library due to missing layout()
2022-09-18 04:21:35 +01:00
ZeroCool940711
997eb12733
Basic implementation for the Concept Library tab made by cloning the Home tab.
2022-09-17 19:02:40 -07:00
ZeroCool940711
cd4396d1da
Added option to enable enable_attention_slicing and enable_minimal_memory_usage, this for now only works on txt2vid which uses diffusers.
2022-09-17 16:49:42 -07:00
ZeroCool940711
c35e5804c2
Added a config option to use embeddings from the huggingface stable diffusion concept library.
2022-09-17 15:25:55 -07:00
Chris Heald
27c8134547
Display the final image before running postprocessing, and don't preview when i=0
2022-09-16 18:25:08 -07:00
Chris Heald
d53643822d
Collect images for final display
...
This is a collection of several changes to enhance image display:
* When using GFPGAN or RealESRGAN, only the final output will be
displayed.
* In batch>1 mode, each final image will be collected into an image grid
for display
* The image is constrained to a reasonable size to ensure that batch
grids of RealESRGAN'd images don't end up spitting out a massive image
that the browser then has to handle.
* Additionally, the progress bar indicator is updated as each image is
post-processed.
2022-09-16 17:39:46 -07:00
ZeroCool
419f3f9474
Merge branch 'dev' into fix-preview-error
2022-09-16 16:36:55 -07:00
Chris Heald
0cf3d1b0ef
When early terminating, generation_callback gets invoked but st.session_state is empty. When this happens, just bail.
2022-09-16 16:29:10 -07:00
ZeroCool
ac99b657d2
Merge pull request #1187 from cheald/enable-batch-preview
...
When running a batch with preview turned on, produce a grid of previe…
2022-09-16 16:27:43 -07:00
Chris Heald
c1ab63c284
When running a batch with preview turned on, produce a grid of preview images
2022-09-16 16:12:52 -07:00
Chris Heald
d22cf56395
Don't attempt mask restoration when there is no mask given ( #1186 )
2022-09-16 23:56:09 +03:00
hlky
b856a91ec8
allow webp uploads to img2img tab #991
2022-09-16 20:55:58 +01:00
bryanlyon
b5462536f8
Fixed decimal prompt weights without leading digit ( #1182 )
...
The regex was not accounting properly for prompt weights that didn't begin with a leading number such as .5 or .1 and was instead splitting those off into their own prompt which got everything all screwed up.
For example, the prompt string of "Fruit:1 grapes:-.5" should parse as
[('Fruit', 1.0), ('grapes', -.5)]
but was being incorrectly parsed as
[('Fruit', 1.0), ('grapes', 1.0), ('-.5', 1.0)]
This fixes that by making the regex properly catch decimals.
2022-09-16 20:42:41 +01:00
