🤝｜tech-support

[ornate elk]

Okay, remove --disable-nan-check and change --no-half to --no-half-vae

[vocal burrow]

is it working now ?

[night orchid]

Ok thank will try

[pearl musk]

Tag Autocomplete: Could not locate model-keyword extension, Lora trigger word completion will be limited to those added through the extra networks menu. [-] ADetailer initialized. version: 24.1.2, num models: 14 ControlNet preprocessor location: /home/void/stable-diffusion-webui/extensions/sd-webui-controlnet/annotator/downloads 2024-02-08 12:04:29,855 - ControlNet - INFO - ControlNet v1.1.440 2024-02-08 12:04:30,005 - ControlNet - INFO - ControlNet v1.1.440 Loading weights [6214e8dd70] from /home/void/stable-diffusion-webui/models/Stable-diffusion/kizukiAnimeHentai_animeHentaiV3.safetensors 2024-02-08 12:04:30,941 - AnimateDiff - INFO - Injecting LCM to UI. /home/void/stable-diffusion-webui/modules/gradio_extensons.py:25: GradioDeprecationWarning: optional` parameter is deprecated, and it has no effect
res = original_IOComponent_init(self, *args, **kwargs)
2024-02-08 12:04:32,466 - AnimateDiff - INFO - Hacking i2i-batch.
2024-02-08 12:04:32,555 - ControlNet - INFO - ControlNet UI callback registered.
Creating model from config: /home/void/stable-diffusion-webui/configs/v1-inference.yaml
Running on local URL: http://0.0.0.0:7860

To create a public link, set share=True in launch().
Startup time: 66.1s (initial startup: 0.1s, prepare environment: 48.2s, import torch: 4.6s, import gradio: 1.8s, setup paths: 3.3s, initialize shared: 0.2s, other imports: 1.6s, setup codeformer: 0.2s, load scripts: 3.7s, create ui: 2.1s, gradio launch: 0.2s).
Applying attention optimization: Doggettx... done.
Model loaded in 8.0s (load weights from disk: 1.8s, create model: 1.2s, apply weights to model: 4.3s, calculate empty prompt: 0.5s).`

No xformers on Optimizations tab

Maybe my GPU is too old for xformers

Bummer

[night orchid]

@ornate elk thanks man, that did it

[pearl musk]

version: v1.7.0  •  python: 3.11.6  •  torch: 2.0.1+cu118  •  xformers: 0.0.20  •  gradio: 3.41.2  •  checkpoint: 6214e8dd70 no xformers still

And it seems I have no way of making a 3.10.6 venv on Arch linux

[marsh arrow]

hi im new here and to stable diffusion, can someone help with something?

i have installed these

but in the dropbox only controlnew is showing

please help i got no clue what im doing

[elder turtle]

Are they installed and setup properly?

[lost crescent]

I have never had a satisfactory answer to the slow load times when changing checkpoints in A1111; when in ComfyUI changes checkpoints almost instantaneously?!

[unreal valley]

can anyone help with this

I am missing models for the clip vision

where should I download them into

[ornate elk]

BasicSR, GFPGAN, Rope and facelib are all no real extensions to use in SD.
These are model files and librarys needed by other extensions and get downloaded automaticly when needed

also roop is outdated

reactor is better

[marsh arrow]

so should i just try installing that?

[ornate elk]

idk what your goal is

if you just want a faceswap extension then yes

[marsh arrow]

idk what my goal is 2 lol

[ornate elk]

then why did you installed all this stuff?

[marsh arrow]

i just wanted to try using SD and read some stuff and though maybe i should start here 😛

like swapping my face for profile pic or something

[ornate elk]

what did you read haha 😄
but yea you should delete the basicsr, gfpgan, rope, and facelib and roop from the extensions folder.
then follow this install guide (read it 2 times) for Reactor Faceswap:
https://github.com/Gourieff/sd-webui-reactor?tab=readme-ov-file#installation
For questions feel free to ask.

[sonic hill]

Is anybody here a bit more proficient with Kohya_ss?

I've been going through every issue on github, but I keep receiving "returned non-zero exit status 1", and none of what is offered as a solution there worked for me

(apologies if this is the wrong server / channel for that)

[marsh arrow]

i dont even remember what ive read its been days now XD

[kindred pumice]

whats the difference between that and standard medvram

[ornate elk]

--medvram is for gpus with 6gb vram or less
--medvram-sdxl is for 8-11gb vram
with --medvram-sdxl only the sdxl models will get loaded with the medvram method and not 1.5.
with medvram everything gets loaded the same

[kindred pumice]

so if I only have the sdxl model than either is fine?

i have 6 gb vram though so probably standard medvram rigght?

[ornate elk]

yes --medvram will be the best for your card

[ornate elk]

and --xformers for much better performance

[kindred pumice]

yup my speed went from 17 to 10 seconds using those both!!

[ornate elk]

if you use 2gb models and euler a as sampler it can be also faster

[kindred pumice]

2gb models? where do I get those

would the quality be less though?

[ornate elk]

nope

[ornate elk]

on civitai.com

dreamshaper v8, or realistic vision v6 for example

[kindred pumice]

which of those is better for photo like images?

[ornate elk]

realistic vision

or analog madness

[kindred pumice]

i'll check that out, thanks!

this dreamshaper 2gb is FAST

about 5 IT/s

actually 7 its lol, it was just wamring up haha

[ornate elk]

nice 😄

[kindred pumice]

only problem rn is that its producing images that look like a painting more than a photo...

is there any way to fix that...

[sour geyser]

happens that differnt browser doesnt post "1006" in commandline.
i noticed that the info it sends to img2img differ to what i see in my main browser, eg: changed prompt/s, automatically loaded picture

[ornate elk]

do you have an example?

[kindred pumice]

cant post images here though

wait nvm its enabled here

they are all like paintings

@ornate elk

[ornate elk]

Sry was afk

[ornate elk]

Whats your prompt?

[kindred pumice]

photorealistic, rows of tulips growing in a field, red yellow purple

tried "photo of" "real looking" and just plain too

[ornate elk]

Is that with Dreamshaper?

[kindred pumice]

yes

[ornate elk]

What's your negative prompt?

[kindred pumice]

none

it looks better but still a painting

let me try something else and see if it doesnt make a painting

wow the city looks pretty good

if I just say "tulips" it is fine apparently

as soon as I add "row" it becomes fake

this is so weird, even if I say "array" it still fakes out

hmm which sampling method should I use?

[fervent bobcat]

Is DreamStudio down?

[kindred pumice]

everything else works fine looks like its only tulips that are broken haha

[dense sandal]

I am trying to use automatic1111

But first time I open it gives me this

I had to delete python 3.11 and reinstalled python 3.10 and reinstalled pip and added the PATHs

But I still have error

Ok after many trials I just seem to fix SOMETHING at least....

First time it opened it gave error in GUI so I downgraded python as suggested

I HOPE now it works good

[night orchid]

Yesterday I had the same problem

I removed all python installations, and then installed Puthon 3.10

everything works now

[tardy sparrow]

Do somebody actually know how to use huggingface effectively ?? i download a model on it (which i guess is for 1.5, but not sure) there is no explanation whatsoever about how which sampling method i have to use or anything

[plain hawk]

Someone, pls help me! I updated controlnet to version 1.1.440 and now it doesn't work😭 How to fix it?

[formal trout]

When I generate on img2img, I get a .jpg and .png

[ornate elk]

can you try an other prompt, like cute cat for example?

it seems like your also need a vae file for better colors

[ornate elk]

then you just have to try out

euler a is a good overall sampler

[kindred pumice]

ok, one sec

[distant plover]

Wow...had SadTalker working, and now it's not working

[nocturne topaz]

Was sad talker, is now just sad 😛

[distant plover]

Im trying this wav2lip

[ancient plover]

I just started using SD (auto 11 webui for amd), does anyone know which option I should use for the "directml memory stats provider" setting in optimizations, if I have a rx 6900 xt?

[ornate elk]

whats in your webui-user.bat?

[stoic cave]

Maybe you should try StableSwarmUI? 1 click install. inless you were wanting Auto for something specific?

[ancient plover]

'--use-directml --opt-sub-quad-attention --opt-split-attention-v1 --no-half-vae --upcast-sampling' currently, although I might change it, since I have been getting errors upscaling (running out of VRAM) with txt2img (not massive upscales either). I saw some posts that claim to fix the issue by adding some more terms...

[ornate elk]

these are the best arguments for amd gpus

its normal that you run out of vram when upscale to high, because directml is not vram optimised

[ancient plover]

ok, so which setting should I choose, or is there a difference?

[ornate elk]

for hires fix?

hires steps: 10
and the resolution can be 512x768 for example

[ancient plover]

no, just what setting is best

[ornate elk]

what setting do you mean?

[ancient plover]

"directml memory stats provider" setting in optimizations

[ornate elk]

performance counter

had no problems with it but didnt tried the other one

i dont think it will make a difference in vram usage at all (but i will test it)

[ancient plover]

ok, thanks, Guess I will have to stick to the extras tab upscaling, since that seems to have no issues with vram...

[ornate elk]

extras tab isnt as good as using hires fix or img2img upscaling.
you only need to adjust the resolution, then you can use hires fix

and use 10 hires steps

for example 512x768 upscaled by 2 will work 100%

then you can further upscale that image in img2img without vram issues

[magic dew]

help

[magic dew]

@sonic hill

[sonic hill]

This doesn’t tell me anything either, what are you trying to do?

[formal trout]

I'm having trouble running https://github.com/Interpause/auto-sd-krita I'm trying to uninstall and re-install Krita on Ubuntu but now everytime I opened I got the error

[magic dew]

I want to use extra photoshop with an old version of stable-diffusion - it was working

[nocturne topaz]

go to the automatic1111 github, find the hash for the older version then open a cmd window and go to the stable-diffusion-webui folder and then use git checkout #hashcode to change the version

[nocturne topaz]

don't use interpause's version..

there is a newer fork that is up to date

#💬｜general-chat message

uninstall old extension completely, then install that one

[magic dew]

I'll test here

[kindred pumice]

dog is super good with this dreamshaper actually

apparently only tulips are broken

let me try some other flower, mabe sunflowers

excellent lol

[tepid crypt]

how worth do you think it is? someone knows what that lora does exactly? i dont understand too complex english terms and many times the translator doesnt translate correctly
https://www.youtube.com/watch?v=kVrd_VDAtUM
and this
https://www.youtube.com/watch?v=MfS3P3AcfsY

[magic dew]

before and after with this error I can't restore the photo

[formal trout]

oh ok

[magic dew]

I tried to install the old version but it gave me this error

[nocturne topaz]

delete the venv folder and try again

[formal trout]

@nocturne topaz I git pull the extension into the extension folder instead of the other, and got

@nocturne topaz I opened Automatic and I don't see the tab, how to install?

[nocturne topaz]

you have to completely remove any traces of the old one first, restart, then install the extension from within a1111, restart a1111 then follow the instructions on the tab in a1111

[magic dew]

same error

[magic dew]

??

[nocturne topaz]

That's not you.. thats for calcu

[nocturne topaz]

open a new command window, go to the venv folder, type activate and press enter - you should see (venv) appear at the start of the file location - then type pip install --upgrade gradio

then try a1111 again

[magic dew]

I've done everything and nothing

[formal trout]

okkk I'll give a try

[distant plover]

well, wav2lip doesn't work. INstalled the extension, but the tab doesn't appear in A1111

[nocturne topaz]

need to use command prompt, not powershell

[nocturne topaz]

is there an error in the console/command window?

[distant plover]

let me look

[magic dew]

It just opens how can I open it?

[distant plover]

none that I can see

Launching Web UI with arguments:
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
Style database not found: C:\Users\Frank\Auto 1111\stable-diffusion-webui\styles.csv
ControlNet preprocessor location: C:\Users\Frank\Auto 1111\stable-diffusion-webui\extensions\sd-webui-controlnet\annotator\downloads
2024-02-08 16:44:30,007 - ControlNet - INFO - ControlNet v1.1.440
2024-02-08 16:44:30,110 - ControlNet - INFO - ControlNet v1.1.440
Loading weights [8a8efa5ad2] from C:\Users\Frank\Auto 1111\stable-diffusion-webui\models\Stable-diffusion\albedobaseXL_v12.safetensors
2024-02-08 16:44:30,473 - ControlNet - INFO - ControlNet UI callback registered.
Creating model from config: C:\Users\Frank\Auto 1111\stable-diffusion-webui\repositories\generative-models\configs\inference\sd_xl_base.yaml
2024-02-08 16:44:31,719 - ControlNetFastload - INFO - Load Controlnet Fastload Filter on isRemote=None and accessLevel=2
Running on local URL: http://127.0.0.1:7860

To create a public link, set share=True in launch().
Startup time: 8.5s (prepare environment: 1.9s, import torch: 2.1s, import gradio: 0.5s, setup paths: 0.5s, initialize shared: 0.2s, other imports: 0.4s, setup codeformer: 0.2s, load scripts: 1.0s, create ui: 1.5s, gradio launch: 0.2s).
Applying attention optimization: Doggettx... done.
Model loaded in 4.7s (load weights from disk: 1.0s, create model: 0.8s, apply weights to model: 2.6s, calculate empty prompt: 0.1s).
Python 3.10.10 (tags/v3.10.10:aad5f6a, Feb 7 2023, 17:20:36) [MSC v.1929 64 bit (AMD64)]
Version: v1.7.0
Commit hash: cf2772fab0af5573da775e7437e6acdca424f26e

Launching Web UI with arguments:
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
Style database not found: C:\Users\Frank\Auto 1111\stable-diffusion-webui\styles.csv
ControlNet preprocessor location: C:\Users\Frank\Auto 1111\stable-diffusion-webui\extensions\sd-webui-controlnet\annotator\downloads
2024-02-08 16:46:27,868 - ControlNet - INFO - ControlNet v1.1.440
2024-02-08 16:46:27,939 - ControlNet - INFO - ControlNet v1.1.440
Loading weights [8a8efa5ad2] from C:\Users\Frank\Auto 1111\stable-diffusion-webui\models\Stable-diffusion\albedobaseXL_v12.safetensors
2024-02-08 16:46:28,235 - ControlNet - INFO - ControlNet UI callback registered.
Creating model from config: C:\Users\Frank\Auto 1111\stable-diffusion-webui\repositories\generative-models\configs\inference\sd_xl_base.yaml
2024-02-08 16:46:29,370 - ControlNetFastload - INFO - Load Controlnet Fastload Filter on isRemote=None and accessLevel=2
Running on local URL: http://127.0.0.1:7860

To create a public link, set share=True in launch().
Startup time: 8.3s (prepare environment: 1.9s, import torch: 2.1s, import gradio: 0.5s, setup paths: 0.5s, initialize shared: 0.2s, other imports: 0.4s, setup codeformer: 0.2s, load scripts: 0.9s, create ui: 1.4s, gradio launch: 0.1s, add APIs: 0.1s).
Applying attention optimization: Doggettx... done.
Model loaded in 4.6s (load weights from disk: 0.9s, create model: 0.8s, apply weights to model: 2.6s, calculate empty prompt: 0.1s).

[nocturne topaz]

press windows key and R, then type cmd and press enter

[nocturne topaz]

no idea sorry - i'd check to see that there isn't --skip-install in the commandline args in webui_user.bat, but otherwise I'm out of ideas...

[distant plover]

ok, thanks

[magic dew]

like this

[nocturne topaz]

try activate.bat then

instead of just activate

ah wait..

need to go to the venv\Scripts folder, then try activate

my bad

[magic dew]

and now

[nocturne topaz]

pip install --upgrade gradio

[magic dew]

erro

[nocturne topaz]

pip install --upgrade gradio_client

[magic dew]

[nocturne topaz]

python -m pip install --upgrade gradio_client

[magic dew]

this is loading

[nocturne topaz]

cool

try that in a1111, and if that doesn't work, try python -m pip install gradio_client==0.8.1

[magic dew]

and now

[nocturne topaz]

try python -m pip install gradio_client==0.8.1

since that didn't uninstall or reinstall anything

[magic dew]

is loading

and now

[nocturne topaz]

now close that and load a1111

[magic dew]

erro

[nocturne topaz]

don't know about that one sorry.. does it work without the --api flag

[magic dew]

without --api

[nocturne topaz]

same error.. no idea then sorry

[magic dew]

I'll be out of work for a few days but thank you

[ornate elk]

Why do you want to use 1.5.0 ?

[ornate elk]

Socket options error can be fixed when adding
httpx
To the requirements.txt

[magic dew]

I really want it because he has it here

[ornate elk]

LCM lora is okay, some models already include them.
But TensorRT isnt recommended as it doesn't support everything in auto1111

[ornate elk]

But that should be in the latest version too

[magic dew]

before it was like this codeformer

When I did this in Photoshop it gave me an error before it didn't give me an error

help me fix this error

[ornate elk]

Have you tried deleting the venv folder?

And let it rebuild

[magic dew]

already

ok

[magic dew]

I just want to fix this error here

[ornate elk]

What did you changed that its not working anymore?

[magic dew]

I formatted my PC, installed everything, when I went to reset the photo I got this error

[nocturne topaz]

I'm guessing new version of extension got installed...

trying to use it with an old version of a1111 didn't work.. all yours @ornate elk 😛

[ornate elk]

I'm off now 😴

[magic dew]

I will try to fix this error

I use this extras page

[timber oak]

hey

how do i vae

i have sd_vae in my ui settings

and equipped "clear vae" in vae settings

cant find it in ui

[formal trout]

I tried but still

[nocturne topaz]

in the pykrita folder you need to put the new files in from the new fork, rather than the old files

i downloaded the github code as zip, unzipped and manually copied the files in to make sure

[formal trout]

here? the fork has files seems to have files for the extension for the Automatic Extension folder

[nocturne topaz]

yep.. replace those files/symlinks with the files from the zip

if that doesn't work.. I'm out of ideas sorry

[formal trout]

I open krita again and now it didn't show error

[nocturne topaz]

win 🙂

[formal trout]

yes but I'm using using --listen and url is http://0.0.0.0:7860/ as I got some error when opening without it

[nocturne topaz]

listen means that anyone that hits port 7860 can use it, whereas 127.0.0.1 is only your computer

[formal trout]

I'm lost on all those dockers

[nocturne topaz]

You can merge them so they form tabs across the top

so you don't have them all visible at once

[formal trout]

well it worked without --listen

[magic dew]

Does anyone know what "string" is?

[nocturne topaz]

string usually refers to a set of characters

i.e. this sentence is a string

[magic dew]

ok

[formal trout]

@nocturne topaz well I got a whole bunch of errors in the console but it seems to generate, but generations doesn't show in the layers

[nocturne topaz]

the inpaint is there with the seed?

[magic dew]

discover the error

[nocturne topaz]

or have you done something other than that?

[magic dew]

what is giving error is

gfpgan

[formal trout]

no idea what

[nocturne topaz]

That's normally the layer that has the generated info, and toggling the eye to the left lets the old version peek through??

or at least that's how I've been using it

[magic dew]

[formal trout]

ah that, it seems that there is nothing on the layers

[nocturne topaz]

try using img2img instead of the inpaint then

you still need the mask from inpaint tab, but you can choose the whole image (ie. make mask and just invert it and you are done) or paint the mask on yourself

[formal trout]

yes that works, but the results

thanks @nocturne topaz for the patiente

[magic dew]

I finished but 1 photo

[nocturne topaz]

there was a tip about creating a feather around the inpaint mask to make it better

it's on the youtube video

see if I can find it from muscle memory

Under select there is one called "Feather selection" - go 20 px and press enter

Shift + F6 is the shortcut

[formal trout]

oh yes it lacked mask blur... but I'm kind of iffy not having that much control on the inpainting, in Automatic I usually play with only masked, padding and I inpaint in full SD XL resolution, here I don't know

[nocturne topaz]

so only masked is default.. the padding is how much Feather selection you give it, and it will choose a good resolution for inpainting

[sonic ferry]

Is this right? Can't mix and match Samplers in ComfyUI or else it's like looking through textured glass?

[lost chasm]

are there any SDXL optimizations i might have forgotten to update or install that can help boost feneration speed?, its a bit slow, like with adetailer/1280x768 and highrez fix set to 1.5x i'm getting gens that take over 3 mins per picture on a 3090, feels emotionally like it shouldnt take this long

[sonic ferry]

Nice, LOL. Took me a sec to figure out what was going on with the arm, I see what it was supposed to be.

[marsh arrow]

dude what did u use and what promts? this looks amazing btw

[novel pawn]

Does anyone know if dreamstudio is broken? Thousands of credits left and it's just not working.

[ornate elk]

You should have a dropdown like the model dropdown

[timber oak]

do u have a screenshot for me?

[ornate elk]

No not at the PC rn

Did you hit apply after adding SD_vae ?

[timber oak]

ok im also still in bed (i didnt sleep tonight)

[timber oak]

i did

i selected the right vae and then activated sd_vae

only those 2

but i cant see them

i even reloaded the ui

maybe i have to restart the whole webui?

[vast hull]

Can someone please help me ? Im getting a cudacnnd.dll error

😭😭😭😭

@ornate elk hi how are u

[ornate elk]

Yes try a full sd restart

[timber oak]

huh

[ornate elk]

Hey, I'm AFK now but can help in 20 mins

[timber oak]

oh

i didnt read the cudapart

i was thinking "what is that file"

[timber oak]

what is the error exactly

[vast hull]

@timber oak

@ornate elk

[timber oak]

oh

[timber oak]

happens when unstart up sd?

[magic dew]

Does anyone know why GFPGAN isn't getting it?

[copper cargo]

trying to use amd

everything got working

but now onnx isnt an argument?

not sure how thats possible

[ornate elk]

The argument got removed and is now a setting.
But I would not recommend to use onnx and olive at all.
You will have a lot of incompatibility with it.
Better use these arguments:
--use-directml --medvram --opt-sub-quad-attention --opt-split-attention-v1 --upcast-sampling --no-half-vae

[ornate elk]

What's your GPU?

[copper cargo]

problem is im getting 2it/s with that. Few months ago before everything broke you could get at least 10 on a 7800xt

[ornate elk]

Yes with onnx and olive.
If you really want to use it, here is the new guide from the dev to get it work.
But if you just want speed I recommend using the Shark webui as its faster than olive but not as restricted.
Both are still worse supported than directml.
https://github.com/lshqqytiger/stable-diffusion-webui-directml/discussions/149#discussioncomment-8392257

[craggy scaffold]

Same issue… haven’t heard anything

[ornate elk]

What did you try?

[magic dew]

use this tool

[ornate elk]

Does it work with codeformer?

[magic dew]

Do you know how to solve this problem?

[magic dew]

?

[ornate elk]

Yes if you answer my question from above xD

If it works with codeformer the next step is to check your models/gfpgan folder

[magic dew]

yes it works with codeformer

So how do I solve it, I need some command

[ornate elk]

Do you have a file in the models/gfpgan folder?

[magic dew]

yes

[ornate elk]

Okay is it a .pth?

[ornate elk]

Okay seems like your missing the gfpgan model

[little swallow]

#1080946152318443610 A Chinese boy, handsome and carefree

[nocturne topaz]

#1047610792226340935

[magic dew]

what do you mean losing

[magic dew]

?

[ornate elk]

Download the file from here and put it into models/gfpgan folder
#🤝｜tech-support message

Then restart the webui

[magic dew]

eeeeeee

Thank you very much man I solved everything now

[ornate elk]

No problem 🙂

[magic dew]

I'm going back to work tomorrow after 3 days

[ornate elk]

Okay

[vast hull]

3090 with 24gb vram

[marsh arrow]

hey @ornate elk what extensions should i use to get these result?

[tight sinew]

I can't get my preview for controlnet to display anything, it doesn't work in txt2img or img2img and there is no error it's just not displaying anything in the preprocessor preview nor does the controlnet image tend to do anything whatsoever

[vast hull]

No it happened qhen installing facegeb

Facegen

[timber oak]

ah

[tight sinew]

anyone know what I'm doing wrong? I've resolved the errors I was getting (I don't think they were related, but I fixed em anyway) but still it won't work 😦

[vast hull]

Yah i have a very expensive gpu like wtf @timber oak

[tight sinew]

I tried different browsers too btw it doesn't display in any of them

[ornate elk]

Make sure the webui is whitelisted in any browser adblocker

Also do you have the models + their config files?

[ornate elk]

Maybe controlnet IP-Adapter can help

[edgy plinth]

Hi

Is dream studio down???

keep getting this

[tight sinew]

I disabled all adblockers and tested on firefox, opera, and edge and all give me no output

[tight sinew]

I do

yay I found it, run preprocessor was what I was missing

ty for the help

[ornate elk]

Oh so you didn't clicked the explosion icon?

[tight sinew]

yes, I didn't know it was an important step and for some reason no youtube videos I watched said to do that

[ornate elk]

Oh okay xD

Yea YouTube tutorials can miss stuff

Also dont watch to old ones. Ai stuff is changing fast

[timber oak]

it seems weird

like

its nvidea right?

i have a 4070

[vast hull]

yeah nvidia

@ornate elk i have a 3090 24gb of vram

[ornate elk]

Okay, and where did you found the facegen program?

I can't find any Programm other than a paid one

[mortal zephyr]

getting a "issue is on our end,try again later" message
been getting that for a full day now
happens to anything and everything i try to create

[fair oxide]

drean studio and API are down!

[half linden]

Can I find a list of the arguments I can add to the .bat file?
I seem to run into some problems with the new version 1,7

[nocturne topaz]

https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Command-Line-Arguments-and-Settings

[ornate elk]

What's your GPU?

[half linden]

nah, I was used to a GUI popping up after start - it would let me add commands, force updates, etc... that is .. gone, it seems

[half linden]

Thank you!

[half linden]

It still works, I am just getting used to the new interface

[ornate elk]

--autolaunch --update-check --update-all-extensions

[ornate elk]

Okay, but if you tell me your GPU I can tell you the best arguments for performance

[half linden]

ah

Thank you.. lemme check so I get it rightr

geforce GTX 1080 Ti

It is not just me who is old 😄

[ornate elk]

Okay then
--xformers --medvram-sdxl --no-half-vae
Should be used

[nocturne topaz]

the 1080Ti is a beast though

way before it's time!

[half linden]

yeah, I am still happy with it... but my rig is 8years old now 😄

another stupid question ... some models have "trigger words", and loras need specific terms at times.
Is there a way to see those easily, or will I have to write them down from civitAI?

[ornate elk]

I think there is an extension like civitai browser that can show them

[nocturne topaz]

Civitai Browser +

[mortal zephyr]

Using image to image SD 1.6 rn and been waiting for a while for it to generate. Am I doing something wrong

[half linden]

ok, so write it down

[ornate elk]

I think the API has some problems as a few people have issues today

[nocturne topaz]

so Civitai Browser + is an extension for a1111 that allows you to search/view/download models from Civitai within a1111

https://github.com/BlafKing/sd-civitai-browser-plus

[mortal zephyr]

Such sad

[nocturne topaz]

Is sd1.6 any good?

I've not tried it

[fair oxide]

their S3 Buckets is full: RpcError: operation error S3: PutObject, https response error StatusCode: 403, api error QuotaExceeded: UnknownError. Does anyone know how to reach to them?

[nocturne topaz]

Yeah, that's a pretty obvious fix

[mortal zephyr]

What is the paid membership for on stability.ai? Like the $20 per month membership vs the credits

[nocturne topaz]

for corporate use of the base model

the credits are seperate

[vocal burrow]

Please note that civitai browser is basically just asking servers "yo about this model what are trigger words that the author registered in the description page". It is not deducing the trigger word from the actual model/lora/etc. So it can fall appart. @mortal zephyr
Also you re not using SD 1.6 but automatic1111's stable-diffusion-webui v1.6 (which is outdated)

[elder turtle]

Contact the support, we cant do much here unfortunately

[fair oxide]

can't find any support mail! can someone provide it to me pls?

[vocal burrow]

https://dreamstudio.ai/support
https://platform.stability.ai/support

[mortal zephyr]

Well looks like I got a lot of things to learn haha. Thanks

[dense sandal]

is there a way to use less vram when using vid2vid? I am using the beta sdxl model

I have 12 vram and it runs out of memory when trying 1024 x 1024, but I can do it with 768x768

[ornate elk]

Do you use --xformers ?

[dense sandal]

it seems so, though it has a bug but I didn't have bug, only out of memory

[fair oxide]

everything seems to work again 🙂

[sleek nacelle]

Right so I tried to move my originally git cloned A1111 webui to an external drive to clear up some space

Found that it was working fine

Until I deleted the old folder for it

Pretty sure it was referencing that folder rather than the new one

And now it's just dumping this error on me

[sonic hill]

Can anybody here help me with Kohya_ss on macOS?

[wind sun]

anyone knows why new folders are not created and all of my new pics deletes

[sonic hill]

Or point me somewhere where I can receive help for Kohya_ss?

[nocturne topaz]

I didnt think kohya was mac compatible…

[sleek nacelle]

I decided to reclone it

[sonic hill]

It is supposed to be! They even have an installation guide for it.

And I get it close to working but get some error that I do not understand:

RuntimeError: Failed to import diffusers.models.autoencoders.autoencoder_kl because of the following error (look up to see its traceback):
module 'torch' has no attribute 'compiler'

[lapis merlin]

so I got comfyui working with directml but the device says privateuseone and only 1gb ram allocated how do I use my gpu?

[ornate elk]

this error appears if the drive isnt formatted to ntfs or wasnt formatted on the pc you use

[nocturne topaz]

Which version of torch is installed?

[sonic hill]

2.0.1

[nocturne topaz]

Is that systemwide or in the venv specific to kohya?

Torch 2 or higher should have the compiler

[steady mountain]

Anyone want to give me a few pointers in setting up a systemd on Ubuntu to try to deal with the memory leak? my system runs out of ram and becomes unstable fast, and I have heard libtcmalloc-minimal4 can help.

[snow lance]

anyone getting these things recently?

started after most recent comfyui update and nodes update

a refresh makes it go away and the red boxes go back to their normal gray

seems random

[sonic hill]

I was unsure so I just ran the command "pip3 install torch torchvision torchaudio" to install the newest version (2.2.0) again.

I still receive the same error, however

[nocturne topaz]

no idea sorry

[sonic hill]

Wait, sorry. My error has actually changed

ValueError: torch.cuda.is_available() should be True but is False. xformers' memory efficient attention is only available for GPU

I have heard someone suggest deleting xformers, but it just reinstalled on startup

He did say to remove "the --xformers flag. --mem_eff_attn"; but I don't know where to remove that from

[nocturne topaz]

not sure in comfy - where do you put the command line arguments?

[sonic hill]

I'm having these issues in kohya, not comfy. And that is exactly the question that I'm asking myself.

[nocturne topaz]

oh.. are you using the command line or bmaltais gui?

[sonic hill]

bmaltais gui

[nocturne topaz]

there will be a toggle box somewhere on the page for xformers iirc

what the gui is doing is collecting all of the arguments together and then when you click train it puts them together in the command so you don't have to type that out.. the --xformers is a flag that usually gets buried in the command with the other settings

so there should be something in the gui that allows you to turn that off

or at least I'm pretty sure I remember it being there

[sonic hill]

That's what I would have thought, but if there is, I can't seem to find it.

You would expect it to be under the parameters right?

[nocturne topaz]

Is there a "optimisations" options under parameters?

[sonic hill]

Yes, that does not contain xformers

I FOUND IT!

[nocturne topaz]

is there something about attention?

[sonic hill]

Paramters -> Advanced -> Weights -> CrossAttention

Let me test with it set to none

RuntimeError: User specified an unsupported autocast device_type 'mps'

[steady thistle]

Did you check your settings under Saving Images, Paths for saving? There are a bunch of directory saving options.

[maiden haven]

Would I regret not using deepspeed?

i guess thats the whole part of my training so nvm

[lapis merlin]

I used this tutorial to install but get bad performance how do I fix it?https://www.youtube.com/watch?v=8rB7RqKvU5U

[ornate elk]

whats your gpu and whats your performance?

[lapis merlin]

6700xt and bad performance my pc freezes

[ornate elk]

hm, maybe you should try auto1111 instead, and check if thats freezes too

also make sure you donw use a model that is 6gb

that could be to much

[median laurel]

……

[nocturne topaz]

….

[steady thistle]

Can anyone help with this error? Trying to use IP Adapter https://huggingface.co/h94/IP-Adapter-FaceID/tree/main version 2

[ornate elk]

it says no input image is given

[ornate elk]

make sure you have these files in your models/controlnet folder.

and these in your models/lora folder:

[steady thistle]

I don't have all of those, I'll add. Thanks!

[sleek nacelle]

So how do I set it to be formatted then

[snow swallow]

Hello people, stoped using stable difussion for a week and when I run again I got a error, tried to delete venv and open it again for a force reinstall but now I get this

No module 'xformers'. Proceeding without it.
F:\AI2\stable-diffusion-webui-directml\venv\lib\site-packages\pytorch_lightning\utilities\distributed.py:258: LightningDeprecationWarning: pytorch_lightning.utilities.distributed.rank_zero_only has been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it from pytorch_lightning.utilities instead.
rank_zero_deprecation(
Launching Web UI with arguments: --use-directml --medvram --opt-sub-quad-attention --opt-split-attention-v1 --no-half --upcast-sampling
Traceback (most recent call last):
File "F:\AI2\stable-diffusion-webui-directml\launch.py", line 48, in <module>
main()
File "F:\AI2\stable-diffusion-webui-directml\launch.py", line 44, in main
start()
File "F:\AI2\stable-diffusion-webui-directml\modules\launch_utils.py", line 676, in start
import webui
File "F:\AI2\stable-diffusion-webui-directml\webui.py", line 13, in <module>
initialize.imports()
File "F:\AI2\stable-diffusion-webui-directml\modules\initialize.py", line 34, in imports
shared_init.initialize()
File "F:\AI2\stable-diffusion-webui-directml\modules\shared_init.py", line 28, in initialize
directml_do_hijack()
File "F:\AI2\stable-diffusion-webui-directml\modules\dml_init_.py", line 69, in directml_do_hijack
import modules.dml.hijack # pylint: disable=unused-import
File "F:\AI2\stable-diffusion-webui-directml\modules\dml\hijack_init_.py", line 5, in <module>
import modules.dml.hijack.diffusers
File "F:\AI2\stable-diffusion-webui-directml\modules\dml\hijack\diffusers.py", line 172, in <module>
) -> Union[diffusers.schedulers.scheduling_lcm.LCMSchedulerOutput, Tuple]:
AttributeError: module 'diffusers.schedulers' has no attribute 'scheduling_lcm'. Did you mean: 'scheduling_ddim'?
Pressione qualquer tecla para continuar. . .

the last one is "press any key to continue" that crashes the terminal

[boreal bison]

Hey guys, my instand ID stopped working in a1111

same everything was working an hour ago

and before this error I was getting oversaturated bad pictures

[nocturne topaz]

Are you using a sdxl model?

[boreal bison]

yes

[nocturne topaz]

If so, are you using the fp16fix vae?

[boreal bison]

It was always set to automatic

when it was working and when it wasn't

[nocturne topaz]

Find and download the fp16fix vae for sdxl and that shoul resolve that issue

[boreal bison]

any side effects on using that vae?

[nocturne topaz]

Nope

[boreal bison]

which one of this should I get?

sdxl.vae.safetensors?

[nocturne topaz]

The third one

[boreal bison]

ok

trying

loaded

same error

restarting a1111

didn't work

[nocturne topaz]

Oh its the unet not the vae

[boreal bison]

enablind this

[nocturne topaz]

Unm

Im guessing it works ok without instant id on

[boreal bison]

aha

I was doing 3d on the side using my gpu, could it be the reason it messed something up?

[nocturne topaz]

I would have thought that would cause an out of memory error rather than a nan

I have had issues where the instantid just didnt like the input I had

[boreal bison]

I've used same inputs multiple times

before

it worked

I also tried a lot of different inputs, it just refuses to work

[nocturne topaz]

Yeah i had that

[boreal bison]

haha how did you fix it?

I thought everything is local and no outside source can affect the workflow all of a sudden no?

simple gens are working

[nocturne topaz]

I cant remember 100%. I think I deleted the model, then reinstalled it - i also had to follow a tutorial to reload it and it wanted 2 controlnets for each gen

[boreal bison]

yeah

[maiden haven]

Just to verify your Cuda version in nvidia-SMI needs to be within .1 minor for backwards compatibility? nvidia-smi has 12.3 I need 12.1. Does the gpus output matter if I'm pathing deepspeed to the 12.1 tooklkit path? it prob does

[nocturne topaz]

Also ugraded extension to v440

[maiden haven]

'd like to not have to roll back my gpus game drivers but thats prob not possible

[boreal bison]

now I'm getting this error wth

[nocturne topaz]

I would have thought cross versioning wouldnt be a great idea but thats way out of scope

[nocturne topaz]

Which file was it looking for?

[maiden haven]

you consider 12.1 and 12.3 to be "way out of scope"?

[boreal bison]

I have no idea

[maiden haven]

or do you mean out of scope for this chan?

[nocturne topaz]

Its out of my wheelhouse but someone in here may know

[nocturne topaz]

Can you gen with controlnet off?

[boreal bison]

yes

[nocturne topaz]

Hmm - when Im next on my laptop Ill find the guide I used

[boreal bison]

ok

[nocturne topaz]

Pretty sure it was a FaceID guide, but again the memory is getting fuzzy as I get old

[boreal bison]

haha

[nocturne topaz]

Will tag you if I find it

[boreal bison]

ok thanks

[nocturne topaz]

https://github.com/Mikubill/sd-webui-controlnet/discussions/2589

[boreal bison]

I feel like controlnet is missing from my extensions

[nocturne topaz]

Sd-webui-controlnet is there

[boreal bison]

ok

Do you have this version?

[boreal bison]

This is the installation guide

[sudden wasp]

Hello, does anyone know how to become an authorized user for finetuning on Stability ais API?

[nocturne topaz]

Yeah once Id followed that guide I never had any issues

Id reinstalled*

[boreal bison]

I did that 😩

Just restarted my machine, let me see if it solves anything

[nocturne topaz]

Did you use a png as input or jpg?

[boreal bison]

Coz' it stopped working randomly, I didn't do anything

I used the same pictures I used 100 times before

[nocturne topaz]

Im sorry my memory is shit.. i did have the same thing!!

[boreal bison]

What's this?

I didn't download this originally, should I?

ohh it workeddddd after restarting PC

wth, seems like afterall it was a gpu cuda crash or something maybe

snapppp again!

[nocturne topaz]

Try with a different sdxl model

[boreal bison]

now it's very random

[maiden haven]

man i can fix python all day long but as soon as i hit a .cpp file .. bam im dead in the water lol

[boreal bison]

Do you use comfyui btw?

[nocturne topaz]

A1111

Have dabbled in comfy so understand it but its not my current goto

[boreal bison]

oh ok

[maiden haven]

this is torture, i have to figure out every incorrect / \ and - /f /IC /I /tp etc in this that isnt a flag in Windows vs linux calling deep mind then i have to go through all the code and fix every function that generations this incorrect command..

i dont know anything about C++ and msvc i thought it was supposed to compile and fix that stuff but idk i cant even compile right now so chicken egg maybe

my brain is going numb remembering / vs \ and which goes which

[uncut glacier]

hi im new to this and ive tried to upscale my video but ive been given this error: OSError: broken data stream when reading image file if anyone can help me fix this asap it would be appreciated ty

for reference i tried to follow the steps on this video and it worked until transferring each pic then the error came https://www.youtube.com/watch?v=7EBbZBTGOA0&t=3s

[maiden haven]

does anyone know what waifu thingis coming from i have no idea where it or that purple menu bar are from

[sonic hill]

Does anybody know how to resolve this issue?

[opaque viper]

Question, why do I alwys get Error: Connection Time Out when I switch to some models? No error in the terminal

[ornate elk]

Make sure the webui is whitelisted in any browser adblocker

[ornate elk]

Hey, do you still get that error?

[opaque viper]

Still not working. Sometimes there's a popup that indicates it's a memory issue, but looking at Task Manager, the ram usage never even reaches 90%

[ornate elk]

What's your GPU and what's inside your webui-user.bat?

[opaque viper]

1660S, no arguments in the bat file

I've used ComfyUI but I've been trying out SDForge for the speed (probably should've mentioned that this is SDForge lol)

[ornate elk]

Yes you should have xD

I didn't looked at sdforge so I dont know what it does.
But if you use auto1111
You would need --xformers --medvram --no-half

[opaque viper]

Yeah. I don't think sdforge needs those stuff, since its built in. I think it is an actual system RAM issue. Since it closes my discord/crashes some of my firefox tabs, before crashing itself lol

[ornate elk]

Oh how much RAM do you have?

You may need to increase your Pagefile to not get crashes

[opaque viper]

16 only. For some reason, some models are fine, but other models crashes it. What's pagefile?

[worn frigate]

hi um

[ornate elk]

Check the model sizes. Sdxl models are 6gb and use more resources.
The windows Pagefile is like a virtual RAM. It goes in action if your RAM is at its limit.
Its for caching files.
Increasing it will mostly stop your crashes

[worn frigate]

i have a weird issue. the controlnet is generating images but not the map. for example depth map segmentation map or lines for canny

[ornate elk]

You have an example?

[worn frigate]

its segmentation map and there is no controlling image

with solid colors

for example no pose image for generating pose.

[opaque viper]

Ah, nice. I'll read up on that. Thanks 😄

[worn frigate]

for example this one

it generated the image in the pose but it does not give that colored skeleton

which it should give beside the result

at this place,beside it

these things

[ornate elk]

Do you have the Controlnet models?

[worn frigate]

yes

almost all of them

[ornate elk]

Can you show a full screenshot of your controlnet settings

[worn frigate]

sure

another example with openpose this time

should it export such image as well

[ornate elk]

and what does the cmd shows?

[worn frigate]

annnnnnd............

now i rememeber

i have a firewall...

[ornate elk]

yea https issue here

"connection issue"

[worn frigate]

ye i always forgot because its so tiny

..cus its tinywall.

[ornate elk]

xD

[worn frigate]

😂

i keep in case some malware runs so cmd does not have permission

not much of safety but ig it does something

idk.. i do not give internet permission to most apps. only a few.

[ornate elk]

Thats a good usage. Also it reduces your limited bandwith, so nothing random gets downloaded

[worn frigate]

ye.

[boreal bison]

I stopped getting it randomly

[bronze sable]

i am new to sd and wanted to kknow which version of python works with sd

i am facing this issue and am nott sure whats casuin it

[dense sandal]

Python 3.10.6

Disable all real time antivirus

Add path and allow to all.users during installation

If all users option is not available (greyed out) uninstall python launcher

[bronze sable]

already off

[dense sandal]

(from add or remove softwarews)

[bronze sable]

yyou mean add pythons adress to webui file

[dense sandal]

Also but no

I mean whe installing

There is an option

Add path to something

Also an option to all users

[bronze sable]

should i delette and reinstall

[dense sandal]

Yes

And dd delete python launcher

[bronze sable]

python is not here

[dense sandal]

I mean uninstall it

Add it

Add the path and the scripts folder path too

[bronze sable]

i added these two paths

[dense sandal]

Ok but if it still doesnt work reinstall python and make sure the two option I mentioned are selected

[bronze sable]

okkay

[dense sandal]

Also delete ven folder (subfolder of auto111 webui folder) when retrying install

Env

[bronze sable]

do i need to check mar anything else?

[dense sandal]

Its okay

Copy that path

So you csn add it later if needed

[bronze sable]

ok

i ran the webui.bat file and its downloading torch

[dense sandal]

Good then

[ornate elk]

Never run webui.bat

Always run the webui-user.bat

[bronze sable]

yeah i ran this file

what happened here?

[ornate elk]

connection issue

[bronze sable]

i dont thinkkkk so its a connection issue cuz everything else was running fine

should i run it again

[ornate elk]

could be because of slow internet connection, timeout, or a Strict filter of a ISP or Government that blocks stuff

[ornate elk]

yea try launching it again

[bronze sable]

its gonna do everything again?

[dense sandal]

Dont think so

I mean tje downloaded stuff should be in cache

[bronze sable]

same error again

[dense sandal]

Copy everything and paste it in chatGPT 😂

Idk

Maybe a permission issue?

[bronze sable]

lol. i dont think thatll work

[dense sandal]

I used chatGPT for my inshallation

And it helped a lot

[bronze sable]

wow

[dense sandal]

Just copy paste it see what it says

Unless anyone knows better

[bronze sable]

i tried to install sd lie some onths ago and it didnt work that tttime too, there were some really knowledgeable people in civitai server, but when i went to civitais server now, it wass completely different

the guides and other stuff were all gone

will this cause anyyy issues or am i good?

[dense sandal]

You good

I would imstall xformers though

Depending on whether you gpu supports it

[bronze sable]

i have a 1650 4gb

[ornate elk]

then you need to edit the webui-user.bat and add --xformers --medvram --no-half
to the line COMMANDLINE_ARGS=

then save and relaunch

[worn frigate]

bro thanks for the -xformers and -medvram line

[bronze sable]

can you explain it in potato terms?

[worn frigate]

Stable diffusion works so fast for me now

I was using it without it for so long until you suggested it.

[worn frigate]

xformers boosts performence for nvidia gpus and medvram means your ram is less than 12 gb and more than 4 (i guess)

it adjusts the stable diffusion performence

[dense sandal]

You need to open it in wordpad

[bronze sable]

hhere?

[dense sandal]

And right after COMMANDLINE_ARGS=--xformers

[bronze sable]

ok

[ornate elk]

--xformers --medvram --no-half-vae
And if you get an error when generating change --no-half-vae to --no-half

[dense sandal]

Thats the same file you open stable diffusion? If so then yes

[worn frigate]

what does --nohalfvae mean

and no-vae?

[ornate elk]

some GPUs (like GTX 16xx cards or AMD RX5xxx)can only load models with fp32 (float point precission)
newer models are made in fp16.
--no-half means (use fp16 models as fp32)
--no-half-vae is the same but only used for vae files and not models

[bronze sable]

it installed xformers

[worn frigate]

oh thanks

[bronze sable]

first diffusion ever lol

btw what is styles.csv?

[ornate elk]

it will get created when saving a style

[bronze sable]

oh ok

[vocal burrow]

Basically a "style" is a shortcut for adding multiple words to your current prompt.

[uncut glacier]

can anyone help me i tried using the extras section and then the batch from directory section, i added in the 2 folders of the pics i wanted to upscale and when it got about half way done i got this error : OSError: unrecognized data stream contents when reading image file but when i check all the images they work fine and load correctly

[vocal burrow]

what does the console log says ?

[uncut glacier]

@vocal burrow

[vocal burrow]

the full log please

copy paste the whole thing, put it in a .txt and drop said file in here

(I should really make an extension or PR to make it easier --")

[vocal burrow]

that's still not the entire log there's stuff before that but ok.

there's probably non png files in one of your directory

[uncut glacier]

bare with me let me grab the full log

theres a lot of text

damn it really wont let me copy and paste XD @vocal burrow but all the files are png did it off da vinci resolve looked through them all

[vocal burrow]

Maybe there's some spaces, weird accent, non ascii stuff in some of their names.

[uncut glacier]

theyre labelled picture then the number in the order its in .png

sorry i probs sound like an idiot and it does have a really easy fix been trying since 6 am my time Xd

[vocal burrow]

Can't really help you more than that without a full log or the images you're trying to convert.

Try to isolate the one picture that's making your SD crash and check for oddities in their name, try opening them with different software to see if they're somewhat broken, etc.

[uncut glacier]

these are the images im trying to convert as for the log it literally wont let me copy and paste it

[snow swallow]

Hello people, stoped using stable difussion for a week and when I run again I got a error, tried to delete venv and open it again for a force reinstall but now I get this

[ornate elk]

hey, try to delete the venv folder and restart the webui-user.bat

[snow swallow]

thanks for the answer, after deleting venv and restarting the webui-user.bat I've got this:

[ornate elk]

there is something cut off

[snow swallow]

the last message is "press any key to continue"

[ornate elk]

can you also show the first/top part of the cmd

[ornate elk]

strange

for me its working with the exact same stuff

try delete the modules folder (not models)

then relaunch the webui-user.bat

[snow swallow]

should I try to delete venv again?

[ornate elk]

do you have git pull in your webui-user.bat?

if so then relaunch again

[snow swallow]

I'm not sure

but relaunching it has the same result

[ornate elk]

can you show me whats inside the webui-user.bat?

[snow swallow]

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--use-directml --medvram --opt-sub-quad-attention --opt-split-attention-v1 --no-half --upcast-sampling
git pull
call webui.bat

[ornate elk]

that looks okay

[snow swallow]

I"m on a AMD

[ornate elk]

i know

[ornate elk]

okay then look in the trash bin, and restore the modules folder

its trange that it doesnt get recreated

[snow swallow]

I even thought of a full reinstall but I'm not sure where to start and move forward

[ornate elk]

For a complete reinstall your in the right channel. If you look at the Pinned Messages of this channel, you'll find my install guide for AMD

Make sure to not install it in the same folder again

If that works you can copy your models over

[steady thistle]

I'm having the same issue. I can't get controlnet to find my image I think.

Hmm, maybe I need --no-half?

I'll try that

*** Error running process: D:\AI\stable-diffusion-webui-directml\extensions\sd-webui-controlnet\scripts\controlnet.py
Traceback (most recent call last):
File "D:\AI\stable-diffusion-webui-directml\modules\scripts.py", line 718, in process
script.process(p, *script_args)
File "D:\AI\stable-diffusion-webui-directml\extensions\sd-webui-controlnet\scripts\controlnet.py", line 1143, in process
self.controlnet_hack(p)
File "D:\AI\stable-diffusion-webui-directml\extensions\sd-webui-controlnet\scripts\controlnet.py", line 1128, in controlnet_hack
self.controlnet_main_entry(p)
File "D:\AI\stable-diffusion-webui-directml\extensions\sd-webui-controlnet\scripts\controlnet.py", line 902, in controlnet_main_entry
input_image, resize_mode = Script.choose_input_image(p, unit, idx)
File "D:\AI\stable-diffusion-webui-directml\extensions\sd-webui-controlnet\scripts\controlnet.py", line 684, in choose_input_image
raise ValueError("controlnet is enabled but no input image is given")
ValueError: controlnet is enabled but no input image is given

[wind sun]

anyone knows where i can download 4x_NMKD-Superscale-SP_178000_G hires.fix upscaler

[steady thistle]

I added multiple images, but it doesn't recognize them.

[steady thistle]

https://huggingface.co/uwg/upscaler/blob/main/ESRGAN/4x_NMKD-Superscale-SP_178000_G.pth You mean this? I literally just googled it

[wind sun]

are you sure the imput folder you pasted is the right one where are all of the pictures ?

[steady thistle]

I didn't use a folder, I uploaded each image to the controlnet multi-inputs

[wind sun]

oh idk then

[steady thistle]

Oh, wait, I forgot the lora. duh.

That did nothing

[wind sun]

im struggling aswell

dw

[steady thistle]

Same error, "no input image."

[steady thistle]

So it can't find your upscaler?

[wind sun]

ye

where is my love CS1o to help those two lost souls

[steady thistle]

Are you getting an error or simply can't find it? Can you screenshot?

[ornate elk]

Put the upscaler .PTH file into the models/esrgan folder

[steady thistle]

Oh, CS1o is here, yay us! lol

[wind sun]

WE ARE SAVED

[steady thistle]

No joke, you're a lifesaver for my hobby. lol

[wind sun]

wym hobby, thats my main job 😄

[steady thistle]

I /think/ it's an issue with my preprocessor. I feel like I'm missing something basic.

Should I place those bin files here: D:\AI\stable-diffusion-webui-directml\extensions\sd-webui-controlnet\models ? I have them placed here: D:\AI\stable-diffusion-webui-directml\models\ControlNet

I followed a youtube video and he placed them in \models\ControlNet, so I did that.

[ornate elk]

Can you explain your problem again

[steady thistle]

When I try this, the image isn't recognized. I'll redo the error and give the info there.

[ornate elk]

Does it work with a single image?

[steady thistle]

No, if I turn off enable, it gives the exact same image.

[snow swallow]

Thanks once again. The full reinstall worked, and was way easier that I thought it would, thanks to your tutorial. Thanks a lot. I had to change the webui to include some changes you suggested for inpaint to work, no everythink is in order

[ornate elk]

Perfect, no problem 🙂
Yea for inpaint on AMD, --no-half is needed

You should get the Tiled Diffusion Extension for getting hires fix to work then

[steady thistle]

Okay, I think I see it is a preprocessor issue. It's not finding the .bin for ip-adapter_face_id_plusv2

I can't get it to work for the first version tho either. So that can't be it.

Yeah, I copied every step of the video exactly and I get no impact on image output wether or not controlnet is enabled.

[ornate elk]

Do you have all files?

And the lora in prompt?

[steady thistle]

I do, yes. I accidentally left it out one try, and added it with the same seed and got a different image, but enabling or disabling controlnet gave same outputs

[ornate elk]

What does the cmd shows?

[steady thistle]

I don't know what you mean? The prompt? (btw, there are only 3 preprocessors showing)

[ornate elk]

The console window

That should show if something doesn't work

[steady thistle]

Ah, okay, one moment

I moved sampling steps to 30 and same issue.

[uncut glacier]

can anyone explain what error this means, when i went into the extra options it gave me this error on the side, i saw on reddit you have to have jpegs which is what my pics are but this error came any solutions would be appreciated

[crystal nacelle]

can anyone help; im using win10 and an AMD and after spending a whole day i managed to install stable diffusion no problem. been using it for afew weeks now. suddenly its nuked again and wont work...

[ornate elk]

whats your gpu?

[uncut glacier]

3080

[ornate elk]

whats in your webui-user.bat?
and how much ram do you have?

[steady thistle]

my bet is pip cache purge. lol, that was my issue that CS1o solved for me.

[ornate elk]

can help if you provide information on the gpu and screenshots of the error

[crystal nacelle]

first i got the xformer issue. so i updated pip and reinstalled xformers and optimum . now stable diffusion runs again but wont generate without throwing me a wall of errors

8GB vram

[uncut glacier]

16 gig of ram, and i honestly just downloaded this so idk how to check whats in mu web ui user bat how do i check for you

[crystal nacelle]

ill launch stable diffusion now again

[ornate elk]

xformers is for nvidia gpus, so that wont work with amd

[ornate elk]

remove --disable-nan-check --reinstall-torch
then delete the venv folder
then relaunch the wbeui-user,bat

[crystal nacelle]

okay let me try

[ornate elk]

right click the webui-user.bat and edit

[crystal nacelle]

i get this error usually. will try your recommendation now and relaunch

[ornate elk]

okay for your 3080 with 10gb vram you need to add:
--xformers --medvram-sdxl --no-half-vae
to the line set COMMANDLINE_ARGS=
then save and relaunch the webui-user.bat
Also make sure the input folder and output folder dont contain spaces in the name

[uncut glacier]

just added the line in thank you and ahhh my output folder does have a space in il change that now thank you il let you know if it works

[ornate elk]

yes, let me know, if that dont work you have to adjust your Windows Pagefile settings

[steady thistle]

@ornate elk I'm going to give up on the controlnet. Like some other things, this seems unsolvable for now.

[ornate elk]

ive tested it rn and it workes for me

its solvable

the .bin files need to be in models/controlnet

the loras in models/lora

then update your controlnet extension

[ornate elk]

controlnet got updated on 01.02.2023

so yours is not latest

[steady thistle]

Ooh! You saw I checked for updates.... How do I update it?

[ornate elk]

normaly its check for updates and then relaunch

make sure you dont have --listen or --share in the webui-user.bat

these block extension updates

[crystal nacelle]

its busy "Installing clip
Installing open_clip
Installing requirements for CodeFormer
Installing requirements"
thanks for the help. ill be back if im stuck again

[steady thistle]

Mine says last updated on: sd-webui-controlnet sd-webui-controlnet.git main 416c3450 2024-01-31 21:29:58 latest

[ornate elk]

okay i see the version is the same

should be good then

[steady thistle]

damn, thought we found it.

This is my Lora folder

I have more in there but I have so many I didn't want to spam

[ornate elk]

and your using an 1.5 based model?

[steady thistle]

Yes:
a girl from 1982, lora:ip-adapter-faceid-plusv2_sd15_lora:1
Steps: 30, Sampler: DPM++ 2M Karras, CFG scale: 2, Seed: 2884366118, Size: 512x512, Model hash: 84d76a0328, Model: epicrealism_naturalSinRC1VAE, ControlNet 0: "Module: ip-adapter_face_id_plus, Model: ip-adapter-faceid-plusv2_sd15 [6e14fc1a], Weight: 1, Resize Mode: Crop and Resize, Low Vram: False, Processor Res: 512, Guidance Start: 0.17, Guidance End: 0.87, Pixel Perfect: False, Control Mode: Balanced, Hr Option: Both, Save Detected Map: True", Lora hashes: "ip-adapter-faceid-plusv2_sd15_lora: a95a0f4bdcb9", Version: 1.7.0

[ornate elk]

okay and what error do you get ?

also is your webui updated?

[steady thistle]

I'm not sure on this... I'llcheck the error

*** Error running process: D:\AI\stable-diffusion-webui-directml\extensions\sd-webui-controlnet\scripts\controlnet.py
Traceback (most recent call last):
File "D:\AI\stable-diffusion-webui-directml\modules\scripts.py", line 718, in process
script.process(p, *script_args)
File "D:\AI\stable-diffusion-webui-directml\extensions\sd-webui-controlnet\scripts\controlnet.py", line 1143, in process
self.controlnet_hack(p)
File "D:\AI\stable-diffusion-webui-directml\extensions\sd-webui-controlnet\scripts\controlnet.py", line 1128, in controlnet_hack
self.controlnet_main_entry(p)
File "D:\AI\stable-diffusion-webui-directml\extensions\sd-webui-controlnet\scripts\controlnet.py", line 902, in controlnet_main_entry
input_image, resize_mode = Script.choose_input_image(p, unit, idx)
File "D:\AI\stable-diffusion-webui-directml\extensions\sd-webui-controlnet\scripts\controlnet.py", line 684, in choose_input_image
raise ValueError("controlnet is enabled but no input image is given")
ValueError: controlnet is enabled but no input image is given

[ornate elk]

what image types are the images?

.png or .jpg?

[steady thistle]

.png

[steady thistle]

How do I update the webui?

[ornate elk]

with git pull

in a cmd inside the folder

[steady thistle]

Okay, will do.

[ornate elk]

click in the explorer bar, type cmd hit enter, and type git pull

you may need to delete the venv folder, since the latest update changed something

but first check if you get an error before doing this

[steady thistle]

I just did it...

venv "D:\AI\stable-diffusion-webui-directml\venv\Scripts\Python.exe"
fatal: No names found, cannot describe anything.
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Version: 1.7.0
Commit hash: 03a88617fcd9442313ce2ed7facfecf6cdf72c36
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
D:\AI\stable-diffusion-webui-directml\venv\lib\site-packages\pytorch_lightning\utilities\distributed.py:258: LightningDeprecationWarning: pytorch_lightning.utilities.distributed.rank_zero_only has been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it from pytorch_lightning.utilities instead.
rank_zero_deprecation(
Traceback (most recent call last):
File "D:\AI\stable-diffusion-webui-directml\launch.py", line 48, in <module>
main()
File "D:\AI\stable-diffusion-webui-directml\launch.py", line 39, in main
prepare_environment()
File "D:\AI\stable-diffusion-webui-directml\modules\launch_utils.py", line 641, in prepare_environment
from modules.onnx_impl import initialize_olive
File "D:\AI\stable-diffusion-webui-directml\modules\onnx_impl_init_.py", line 7, in <module>
import optimum.onnxruntime
ModuleNotFoundError: No module named 'optimum'
Press any key to continue . . .

I'll delete the venv next

[formal trout]

After installing ReActor I got some Onnx error thing. Now I got it with ip_adapter_face_id_plus

I'm on Ubuntu, Radeon 6700, Automatic1111

[ornate elk]

Yes that will fix this

[steady thistle]

installing still

it gives me a "restart the server" when I shut down the cmd and restarted. I'll reboot my PC.

actually you tell me what I need to do next?

I did pip cache purge... 430 files deleted so maybe that will help.... it's giving the same error

[lapis merlin]

anyone know why it says my gpu ram is only 1024 ?

[formal trout]

well I'm reinstalling Automatic... a good chnage to get rid of unused stuff

This extension clenas de VRAM after each generation, is it necessary do you think? Generating takes longer as it pauses for each generaation https://github.com/miaoshouai/miaoshouai-assistant

should I do this? Last time I didn't, I think some extension installed

[ornate elk]

Go into the Controlnet settings

There select the load clip model in cpu

[wind sun]

so this problem is still here

my photos are not getting saved after generating

[ancient rune]

The dashboard wonr load and the api seems down

Error code 522 - Timed out

[steady thistle]

Wow, I'm getting results. Thanks!

[steady thistle]

I'm still getting these errors, is there anything I can do about this?

[ancient rune]

am I the only one ? Can someone try to log in his account pls ?

[formal trout]

wow this was misfortune with luck, as I was using a broken or not installed properly instant id, now it works as intented

[ornate elk]

Hm idk but maybe check if the seed mode is set to cpu

[steady thistle]

It wasn't, trying that.

[thin cloak]

Getting API errors here too - can't load dashboard either to view my credits etc.

[steady thistle]

Is this correct?

[ancient rune]

same

err 522

timed out

[craggy scaffold]

same

been having a ton of trouble with dreamstudio recently tbh

it's been frustrating

[steady thistle]

It was working, now it's broken again.

TypeError: 'OnnxRawPipeline' object is not callable

[ornate elk]

No

Don't use anything with olive

Or onnx

[steady thistle]

Thanks, probably my issue. Trying to fix stuff and getting distracted causes errors

Ugh, turned it off and still not getting any results

AttributeError: 'OnnxRawPipeline' object has no attribute 'sd_model_checkpoint'

I can't figure how to turn off onnx

[azure sable]

All of my training images are above 1024 pixels on the smallest side and yet buckets is making them way smaller. Is there a way to fix this besides cropping all my pictures to square resolutions? Would this affect SDXL training in any way?

[ornate elk]

Make sure you revert and don't do anything that is stated here:
https://github.com/lshqqytiger/stable-diffusion-webui-directml/discussions/149#discussioncomment-8392257

[mortal zephyr]

API/Dream studio down?

[maiden haven]

If I put all of my backend py / sd / deepspeed / torch(vision|audio|etc) / dreambooth stuff in wsl2, what sort of gotchas are there? 10-20% perf drop I've read which if it gets me around this deepspeed windows torture is fine. Is there anything else I'm not considering? Are there any nvidia drivers or apps that don't work/are awful in wsl/linux sort of thing? I'm on a 4090 with 128gb ram etc, and i don't need high perf it's just hobby. thanks

i'm having to constantly fix a ton of python all over the place to get this all working on windows and it's probably a sunk cost now but i know linux way better and if that's mostly just conda installing req files etc and i get all of the latest deepspeed / torch / cuda etc then i dont know what i'm missing

[ornate elk]

idk why anyone would prefer running nvidia stuff on linux or even wsl

on windows you will have the best usage of these tools

wouldnt recommend Dreambooth for training as it breaks SD most of the time

Better use Khoya_ss to train

[maiden haven]

i'd prefer running it on wsl/linux because of what i just mentioned, I'm on day 2-3 if trying to get everything working together in windows, i'll fix the pytorch / deepspeed versions then break something else, constantly fixing package deps that don't work together on windows, no AIO, etc

have you done deepspeed / dreambooth on windows? unless i'm going the wrong way it's a PITA

I'm literally walking through errors trying to fix everything that looks important wackamole... i had it working last night but now it's being problematic again

Oh I'll check out khoya thanks

[ornate elk]

I've tried dreambooth one time for embedding training, then switched to Khoya_ss for Lora training

And it supports dreambooth training too

Always Better to have a separate tool for training

[maiden haven]

I'm reading khoyas_ss git and I'm not 100% sure where it goes, is it another frontend specifically pointing to khoya on its on port or is it an sd-webui replacement/theme?

it kinda just says "khoya windows-focused gradio ui" not ure if app or what

Oh separate tool so I guess the former

[steady thistle]

Don't do that.. tricky but will try to undo when I'm back.

[maiden haven]

i feel like the best way to run this would (in my mind, SHOULD) be to throw all of the backend tools/services into docker containers in a docker-compose.yaml and run it on linux but I can't remember if you'd have to VFIO mount the GPU to the SD container or if every container would have direct access, it might need a cgroup tweak.. breaking the sd parts between machines im sure is a thing so i'm assuming most of its API to API but a lot of what i'm working on is clearly directory/config file based

I have an extra 2080 super OC sitting around, I know its not rtx/cudas but would it provide any benefit at all to SD stuff? would it just be like an insanely slow 4090 w/ no cuda + old

[umbral plover]

Question, why does FaceDetailer don't detect face on images that was were upscaled?

[ornate elk]

2080 is cuda and rtx supported

It will work really good for image generation

And its capable of training at low speed

[maiden haven]

oh it is an RTX awesome i forgot

so in that case ive got a whole ryzen 7 5800x + rtx 2080 computer sitting around i could finally use for something

[ornate elk]

Yea thats a good setup for SD

[cinder moth]

I'm having issues with installing the roop, i've downloaded the visual studio and the requriements. But it's not showing on my img2img

I've tried this solution i found on reddit but didn't help as well;

[maiden haven]

it exists prase be nvidia https://github.com/NVIDIA/nvidia-container-toolkit

[maiden haven]

what is the error / what is it doing not doing

[ornate elk]

Roop is outdated, use Reactor instead

But read the installation instructions of it carefully on github

[maiden haven]

Ok, 1:12pm, lets see how quickly i get all this up with containers

[sinful frost]

Just re-installed Auto1111 and can't seem to get the Captioning in the Extras tab to work. Am I missing anything I need to install?

[steady thistle]

I had to set it to CPUExecutionProvider in the Onnx tab. I think that's working right now.

[ornate elk]

Ah okay

[maiden haven]

wow wsl2 doesnt have built in snapshot/backups, just a filesystem export dump

[steady thistle]

And thank you!

[ornate elk]

No problem 🙂

[magic dew]

because roop is not catching

[boreal bison]

can I make the preview updates more frequently?

it does it twice for 25 steps

[ornate elk]

yes in the preview settings

[steady thistle]

It's horribly slow trying to do SDXL Turbo. I have to run --no-half and --lowvram to do a small SDXL image. Sadly, it's so rough on speed I think I'll have to stick with ReActor. I'm still getting some weird Onnx errors in the startup, but it's working at least.

[ornate elk]

yea sdxl on AMD is so slow

[thin cloak]

API still down for everyone else?

[boreal bison]

let me find

[steady thistle]

It's worse than it was. I think this error is killing my vram. Let me post the error, I think something bad is happening. It's taking 10x as long and failing when it hasn't for weeks.

[boreal bison]

worked. it slows down the generation seems like huh

[steady thistle]

Might not be so bad after changing things and restarting. Crossing fingers.

[ornate elk]

yes

each image preview costs vram and time

[boreal bison]

makes sense

[willow aspen]

so my models are still running really slow and i am thinking its some type of memory leak

because i have tried variations of all the different command line args and none of them really make a difference but if i reboot my pc and free the memory the models take up its way faster

it starts fast, but if i keep generating and using different models, RAM is getting slowly taken up and eventually its unusable

[crystal nacelle]

I wanted to ask for some tips on getting quicker renders and generations with an AMD card?
Ive got an RX7600 8GB with these command lines:
"--use-directml --upcast-sampling --lowvram --opt-split-attention"

[ornate elk]

you can add --opt-sub-quad-attention

[maiden haven]

god i feel good being back home on the ole lunix, no hopping from admin to powershell to anaconda terminals and almost all of my packages work, 1 vscode window and im chillin

[snow swallow]

I'm not sure here is the right place, but I'm having trouble with resize. I generated an image than went to img2img - generation img2img tabs and chose resize by 2 (from 512x512 to 1024x1024) but after I click generate some time passes and I get a image, slightly different but still 512x512 not sure what I'm doing wrong

[rugged latch]

boop

how do i make stable use the APU instead of mai cpu

or is that even possible

[ornate elk]

Did you remember the two times I said your webui is running on CPU instead of GPU 😄

[rugged latch]

;^;

[ornate elk]

If you want to run it on GPU you need to follow my install guide for AMD. Its in the pinned messages of this channel.
Make sure to install it into a new folder and do the important step mentioned there too.

If you already have the stable-diffusion-webui-directml folder,
And not the Stable-diffusion-webui
You can skip the start and start at step 5

I'm off for now but if you face any problems, or have questions feel free to write them down here, then I can take a look tomorrow or someone else can help