Commit
…rience
- Loading branch information
There are no files selected for viewing
30 comments
on commit 22bcc7b
There was a problem hiding this comment.
-- UPDATE --
it looks like working now... for whatever reason...
still not working
- latest commit
- on Google Colab and latest Firefox on Windows
- launch option
--share --api --gradio-queue --xformers --medvram --disable-safe-unpickle --enable-insecure-extension-access --no-half-vae --no-download-sd-model --disable-console-progressbars
There was a problem hiding this comment.
yup looks like it's working fine
There was a problem hiding this comment.
This seemed to fix loading txt2img screen correctly, but I'm still getting infinite load in pnginfo tab when using send to txt2img button
There was a problem hiding this comment.
This seemed to fix loading txt2img screen correctly, but I'm still getting infinite load in pnginfo tab when using send to txt2img button
yeah, I got no issue with txt2img after this commit and the infinite spinning logo is gone.
what I'm having issue now is the img2img.
- sometimes it work
- sometimes I have to reload the WebUI and regenerate the image and then I can send it to img2img
- sometimes I clik on generate but didn't see the result on the WebUI and i have to dig into the output folder and manually find the image
- sometimes the generate button just doesn't send any command to the backend.
not really sure about this one, could just be my connection...
There was a problem hiding this comment.
The front-end performance is not good, there are too many Gradio loading animations, and they are too frequent. I don't know if it's caused by some bug, but the user experience is not good anyway. When I reverted to the commit 803d44c, everything was completely normal.
There was a problem hiding this comment.
I'm still getting the hanging but it seems on and off, if I refresh the page sometimes it goes away, sometimes not
There was a problem hiding this comment.
North of 150 seconds to load extensions, 320 so far to check for updates
There was a problem hiding this comment.
lmfao
There was a problem hiding this comment.
- Sometimes webui getting stuck and I need to relaunch it completely - reloading ui do not help
- Output images after generation in txt2img and img2img do not pop up, images just dissapear from ui. But everything is fine in output folder
There was a problem hiding this comment.
After the update, the program became "stupid" and began to give very bad results in img2img and inpain. Is there any way to go back to the previous version or fix the problem?
There was a problem hiding this comment.
Sometimes webui getting stuck and I need to relaunch it completely - reloading ui do not help
- Output images after generation in txt2img and img2img do not pop up, images just dissapear from ui. But everything is fine in output folder
exactly this. glad that I'm not the only one experiencing this
There was a problem hiding this comment.
-- UPDATE -- it looks like working now... for whatever reason...
still not working
- latest commit
- on Google Colab and latest Firefox on Windows
- launch option
--share --api --gradio-queue --xformers --medvram --disable-safe-unpickle --enable-insecure-extension-access --no-half-vae --no-download-sd-model --disable-console-progressbars
remove --gradio-queue，add --no-gradio-queue #9074 (comment)
There was a problem hiding this comment.
Hello
I try you stable-diffusion-webui and install it on my computer on windows 10 it seems to work, at least i try txt2img and it works but for some reason i think the css are not loaded
how can i fix this
There was a problem hiding this comment.
When I reverted to the commit 803d44c, everything was completely normal.
This commit 803d44c fixed the problem for me. I've pushed a Docker Image with this exact commit of stable diffusion webui:
docker pull rubbrband/sd-webui
docker run -it --gpus all rubbrband/sd-webuiI've also built an open source cli tool to automate training diffusion models, it supports stable diffusion webui too!
Install it by running:
pip3 install rubbrband
rubbrband web sd-webui
There was a problem hiding this comment.
@artisanfx related - #8537 (maybe china related)
please continue further discussion in issues, which are more suitable.
There was a problem hiding this comment.
Same problem here...how to fix?
There was a problem hiding this comment.
Same problem here...how to fix?
this commit fixes the issue.
also, use --ngrok {your authtoken} --ngrok-region {choose region} and remove --share.
ngrok gives better performance and stability compared to gradio's CDN.
There was a problem hiding this comment.
This commit still has issues if you're using Firefox; checking for extension updates becomes an infinite loop. Edge is better, but even then it's slower than before March 24 releases.
There was a problem hiding this comment.
the function in script.js get_uiCurrentTab is not uptodate with gradio 3.23.
it always return "txt2img" tab, regardless which tab is really selected.
the correct implementation should be:
return gradioApp().querySelector('#tabs button.selected')
check the screenshot: tab "img2img" is active, and in console you see new and old implementation.
There was a problem hiding this comment.
This commit still has issues if you're using Firefox; checking for extension updates becomes an infinite loop. Edge is better, but even then it's slower than before March 24 releases.
I'm on the latest firefox and I can confirm that there's no infinite loop when checking for extension.
There was a problem hiding this comment.
@ClashSAN
Thanks for the link and yes it was the problem of the register
HKEY_CLASSES_ROOT.css
was somethinks like aplication/x-css
change it to text/css and restart the webui.bat and ctrl/F5 to reload the page and the screen works fine now
thanks a lot
There was a problem hiding this comment.
This commit still has issues if you're using Firefox; checking for extension updates becomes an infinite loop. Edge is better, but even then it's slower than before March 24 releases.
I'm on the latest firefox and I can confirm that there's no infinite loop when checking for extension.
Nice, I wish I had some of that luck!
I tried without any extensions at all and it ran fine, but adding some extensions back in makes it slower. I'm now testing by adding them back in one by one. At first I thought it was the AlUlkesh image browser, but it seems that having any extension throws it for a loop about 10-20% of the time now. When I added the image browser extension back in, it seems more like 40-50%.
Taking out "--no-gradio-queue" from my startup blew it all to hell, the 10-12 second check to look for extension updates without the queue has now crossed 200 seconds and counting :(
Something's messed up with gradio 3.23.0 it seems, and not for a small number of people.
There was a problem hiding this comment.
This commit still has issues if you're using Firefox; checking for extension updates becomes an infinite loop. Edge is better, but even then it's slower than before March 24 releases.
I'm on the latest firefox and I can confirm that there's no infinite loop when checking for extension.
Nice, I wish I had some of that luck!
I tried without any extensions at all and it ran fine, but adding some extensions back in makes it slower. I'm now testing by adding them back in one by one. At first I thought it was the AlUlkesh image browser, but it seems that having any extension throws it for a loop about 10-20% of the time now. When I added the image browser extension back in, it seems more like 40-50%.
Taking out "--no-gradio-queue" from my startup blew it all to hell, the 10-12 second check to look for extension updates without the queue has now crossed 200 seconds and counting :(
Something's messed up with gradio 3.23.0 it seems, and not for a small number of people.
I rebased my colab notebook to use TheLastBen's code and I don't really have any issue using the latest commit.
I even have some extensions and ControlNet installed and it's still working fine.
the only issue is that I can't easily switch between latest commit or previous commit.
this is not related to the topic because I modified his code and now I can't switch between commit without getting traceback and I do know I'm on my own in this one.
There was a problem hiding this comment.
AUTOMATIC1111 solve version conflict problem
Please refer to the manual below for the solution.
please update colab
#ddetailer 무한로딩 해결 방법!! 버전 충돌!! 해결방안!!
딥러닝 torch 버전 변경
pip uninstall -q -y torch torchvision torchtext torchdata torchaudio
pip install -q -U torch==1.13.1+cu117 torchvision==0.14.1+cu117 torchtext torchdata torchaudio --extra-index-url https://download.pytorch.org/whl/cu117
xformers 엔비디아 렌더링 빠르게하는 라이브리 버전 변경
pip uninstall -q -y xformers
pip install -q -U xformers==0.0.16rc425
ddetailer 의존 패키지 미리 설치
pip install -q -U openmim==0.3.7
mim install -q -U mmcv-full==1.7.1
pip install -q -U mmdet==2.28.2
There was a problem hiding this comment.
Install/Update AUTOMATIC1111 repo
insert code here!!
...
.....
with capture.capture_output() as cap:
%cd /content/gdrive/$mainpth/sd/stable-diffusion-webui/
!git reset --hard
time.sleep(1)
!rm webui.sh
!git pull
if Use_Latest_Working_Commit:
!git fetch --unshallow
!git checkout a9eab23
!pip install gradio==3.16.2 --no-deps -qq
else:
!pip install gradio==3.23 --no-deps -qq
## add Paste Version change to use detection detailer 디텍션 디테일러를 사용하기 위한 버전 변경
# 딥러닝 torch 버전 변경 Deep learning torch version change!!
!pip uninstall -q -y torch torchvision torchtext torchdata torchaudio
!pip install -q -U torch==1.13.1+cu117 torchvision==0.14.1+cu117 torchtext torchdata torchaudio --extra-index-url https://download.pytorch.org/whl/cu117
# xformers version change error!! solve the problem
# xformers 엔비디아 렌더링 빠르게하는 라이브리 버전 변경
#!pip uninstall -q -y xformers
#!pip install -q -U xformers==0.0.16rc425
# ddetailer 의존 패키지 미리 설치 ddetailer pre-install dependencies,mmcv-full version change, mmdet version change
!pip install -q -U openmim==0.3.7
!mim install -q -U mmcv-full==1.7.1
!pip install -q -U mmdet==2.28.2
clear_output()
....
AUTOMATIC1111 TO-DO Please put function by version!!!
Deep learning torch version change
xformers version change error!! solve the problem
ddetailer pre-install dependencies
mmcv-full version change
mmdet version change
You have to create an option for each Colab version and change it.
You have to create version-specific options to solve compatibility issues.
There was a problem hiding this comment.
Install/Update AUTOMATIC1111 repo insert code here!! ... ..... with capture.capture_output() as cap: %cd /content/gdrive/$mainpth/sd/stable-diffusion-webui/ !git reset --hard time.sleep(1) !rm webui.sh !git pull if Use_Latest_Working_Commit: !git fetch --unshallow !git checkout a9eab23 !pip install gradio==3.16.2 --no-deps -qq else: !pip install gradio==3.23 --no-deps -qq ## add Paste Version change to use detection detailer 디텍션 디테일러를 사용하기 위한 버전 변경 # 딥러닝 torch 버전 변경 Deep learning torch version change!! !pip uninstall -q -y torch torchvision torchtext torchdata torchaudio !pip install -q -U torch==1.13.1+cu117 torchvision==0.14.1+cu117 torchtext torchdata torchaudio --extra-index-url https://download.pytorch.org/whl/cu117 # xformers version change error!! solve the problem # xformers 엔비디아 렌더링 빠르게하는 라이브리 버전 변경 #!pip uninstall -q -y xformers #!pip install -q -U xformers==0.0.16rc425 # ddetailer 의존 패키지 미리 설치 ddetailer pre-install dependencies,mmcv-full version change, mmdet version change !pip install -q -U openmim==0.3.7 !mim install -q -U mmcv-full==1.7.1 !pip install -q -U mmdet==2.28.2 clear_output() ....
AUTOMATIC1111 TO-DO Please put function by version!!!
Deep learning torch version change xformers version change error!! solve the problem ddetailer pre-install dependencies mmcv-full version change mmdet version change
You have to create an option for each Colab version and change it. You have to create version-specific options to solve compatibility issues.
or you can just use my colab notebook or TheLastBen's notebook
There was a problem hiding this comment.
Little girl
There was a problem hiding this comment.
No
There was a problem hiding this comment.
yup looks like it's working fine
s