🤝｜tech-support

[lusty sage]

Yeah that's the main reason I wanted it for

I tried to set them all to min and still no affect

[ornate elk]

can you restart the webui-user.bat and show me a full cmd log?

[lusty sage]

I will once I'm back home

[rough void]

was is das?

[ornate elk]

you should get like a 18k score for GPU

[ornate elk]

whats your GPU? and whats inside your webui-user,bat?

[rough void]

amd 6950 xt

"webui-user,bat" idk

[ornate elk]

your on windows or linux?

[rough void]

windows

[ornate elk]

the webui-user.bat is the file you always launch to start the webui

right click and edit it

[rough void]

you mean tihs?

[ornate elk]

yea this

and that wont work for AMD

[rough void]

what do i do?

[visual orbit]

Figured, I am seeing results from sellers and trying to see what is best. This guy runs it with a 850w PSU. Is this the problem? The hotspot goes to 90© while at the 20minute mark. What is your opinion on this? Is a user error or is there something wrong with this card?

[ornate elk]

I made a Guide to get Stable Diffusion working with AMD using Zluda.
The guide is in the pinned Tabs of this channel or linked here:
#🤝｜tech-support message
Follow every step exactly and it will work 🙂
For any Questions feel free to ask me. Im on AMD too

[ornate elk]

hmm idk, but could be a bit to hot then. 850w can work

seems like hes testing it in a barebones rig with a low cpu

[shell crystal]

Can someone help me with adetailer? I'm currently using runpod stable diffusion automatic1111 and installed it via the extension tab, but after restarting and enabling it, it appears to do nothing even after double checking to make sure it's enabled

These are the current settings I have

[visual orbit]

Yeah, he is not very knowledgeable about pcs as I have understood. Still, is it possible the card is failing in some way?

[rough void]

is that right?

[ornate elk]

no, you have to delete everything SD related youve downloaded before using my guide to make sure you start correctly

its not done by editing the wrong webui version

you can leave Python 3.10.11 and Git installed

these are needed

[rough void]

what file you want me to del?

this?

[ornate elk]

idk what webui version you downloaded before

[ornate elk]

yea this one is for nvidia only

you can delete it

[rough void]

ok

[ornate elk]

idk, the 3090 got used a lot as mining cards back in the day

[rough void]

this?

and download the new one?

[ornate elk]

yep

[visual orbit]

This was bought in 2022 and has played only some games per sellers word

[ornate elk]

okay

[rough void]

is it loading or?

[ornate elk]

the download is done

you cna do the next steps

[rough void]

ok

[ornate elk]

oh wait

no

it got an error

[ornate elk]

make sure your internet connection is stable, also check that your PC Date/clock time is correct synced, also restarting the PC can help

[rough void]

internet connection is not stable

can i try agien?

[ornate elk]

yep

you can run the command again

[rough void]

did it work?

[ornate elk]

it was at 21%

[ornate elk]

failed again

best is to connect over ethernet, and disabled any VPN services

dont download stuff in the background etc

[rough void]

cant i download it from goggle to pc?

[ornate elk]

you could download it as zip, but the installation process will also need multiple files and more GB, so if your connection is that bad i dont know if it would work

[rough void]

i can try

[ornate elk]

you need to go on the github repo here:
https://github.com/lshqqytiger/stable-diffusion-webui-amdgpu
then click on the Green "Code" Button on the right. Then select "Download Zip"

[rough void]

ok

[ornate elk]

then move that to the sd-zluda folder and unzip it there

[rough void]

still no?

[ornate elk]

nope

[rough void]

still work fine if winrar

[ornate elk]

have you tried restarting the PC ?

[rough void]

it work?

[ornate elk]

thats the downloaded file right?

as zip

[rough void]

yes

no

i extract it

[ornate elk]

yea

[rough void]

now i have done this

now i launch and get error

loading

do i need to put in model agien?

in model folder?

i dident get a error btw

[ornate elk]

only if you already have a model, if not it will download the default model (4.5gb)

[rough void]

can i use this?

https://prompthero.com/ai-models/jh-s-samdoesarts-download

[ornate elk]

yes

you need to move it into models/stable-diffusion

[rough void]

yeah ik how to that

this gonna take i while so ill be back when its done

[lusty sage]

I updated and now I get this error when enabling bmab

"*** Error running postprocess_image: C:\auto1111\sd.webui\webui\extensions\sd-webui-bmab\scripts\sd_webui_bmab.py1it/s]
Traceback (most recent call last):
File "C:\auto1111\sd.webui\webui\modules\scripts.py", line 897, in postprocess_image
script.postprocess_image(p, pp, *script_args)
File "C:\auto1111\sd.webui\webui\extensions\sd-webui-bmab\sd_bmab\bmab.py", line 81, in postprocess_image
p.context.index = self.index
AttributeError: 'NoneType' object has no attribute 'index'"

[quasi ginkgo]

Sup yall, I just updated my extensions and have started getting an error. I can post the full log if necessary but it looks to be mainly this - "RuntimeError: jaxlib version 0.4.28 is newer than and incompatible with jax version 0.4.25. Please update your jax and/or jaxlib packages."

I've checked the issues tab on A1111 and someone posted that they were having this issue and have updated the package by going into their venv and starting the "activate" script, and installing the proper jaxlib version... I don't seem to have a venv folder in my Webui dir, nor could I pinpoint said script. I have done 'pip install jaxlib==0.4.25' but that hasn't anything(Not sure if I even have a virtual environment set up since I did the 1 click A1111 installer, if not - then jaxlib0.4.25 was installed locally and not in the venv right?)

Haven't done a full reinstall, I'm considering that the nuclear option.

[ornate elk]

Hey, if your using Auto1111 you should have a venv folder

[ornate elk]

Maybe you need to reinstall the extension or delete the venv before

[quasi ginkgo]

Should be located in my webui folder yea?

I don't see it

Is there a crash log I can find and post?

[ornate elk]

You can launch the webui and copy the whole cmd log

[ornate elk]

Ah okay, didn't saw that error before. But without a venv its not easy to fix.
Best would be you reinstall Auto1111 but following my guide to install it correctly. Because downloading it as zip will get you in trouble again.

My guide is in the pinned messages of this channel

[quasi ginkgo]

Checked the pins earlier and saw that it was a no-no to have my install inside my Pictures folder so I moved my install to my root C: before I asked for help here 🫢

Would it be problematic to backup my loras, models, text invs, ect... ?

to use my backup*

[ornate elk]

Nope you can backup them and easily drop them back into the new install. But don't backup your extensions folder,
only models, loras, embeddings, outputs, upscalers

[quasi ginkgo]

Can do

thanks boss

[ornate elk]

No problem 🙂

[rough void]

is done?

[zinc dagger]

it wont run now... i install animatediff

[sinful needle]

Animatediff error. Only happens when I enter a certain length of prompt.

[agile perch]

hey

can i get some help?

with automatic1111?

this has been going on for like 11 minutes or more

im not sure if its stuck or actually isntalling in the background

[fair oxide]

i also need help

[agile perch]

pressed esc and it got me through

but idk, hope i didnt miss anything important...

[fair oxide]

with automatic1111 File "C:\AI\sd.webui\webui\repositories\k-diffusion\k_diffusion\evaluation.py", line 6, in <module>
import clip
File "C:\AI\sd.webui\system\python\lib\site-packages\clip_init_.py", line 1, in <module>
from .clip import *
File "C:\AI\sd.webui\system\python\lib\site-packages\clip\clip.py", line 6, in <module>
from pkg_resources import packaging
ImportError: cannot import name 'packaging' from 'pkg_resources' (C:\AI\sd.webui\system\python\lib\site-packages\pkg_resources_init_.py)
Press any key to continue . . .

was trying the windows automatic install ver

[quasi ginkgo]

@agile perch nice pfp

[agile perch]

oooooo

pfp bros :D

[quasi ginkgo]

ayyyyyyyyyyy

[fair oxide]

Hey anyone

mportError: cannot import name 'packaging' from 'pkg_resources'

[lusty sage]

did both of these and still getting the same error, may need to do a clean install

[quasi ginkgo]

nvm path wasnt set, "where python" worked for getting the dir

[hollow silo]

can anyone figure this one out? there is no real rhyme or reason why it works sometimes but not others. I am using ComfyUI, the same settings work in A1111 with no issues.

oops bottom left was supposed to be this

[steel geyser]

looks like the two guys in vc rn are bots

sebastian cortes and nodnolty

[cinder dirge]

hey guys, looks like API is down and none of my calls are working anymore to https://api.stability.ai/v1/generation/stable-diffusion-xl-1024-v1-0/text-to-image
They just timeout afte 30s.

[proven vortex]

hey peeps, i have a problem, im on automatic ui and after updating i have no preview image however it generates the images fine

[ornate elk]

make sure the webui is whitelisted in any browser adblocker, also after updating you can delete the venv to let it recreate a fresh one

[proven vortex]

ooh, what's a venv?.. I'm sorry I am new to this

just delete this whole folder?

[ornate elk]

yep

[proven vortex]

I am getting this now, i deleted the venv folder which was around 8 gigs, then it redownloaded 2.5 gigs when i started it again

[ornate elk]

what does the cmd shows when starting up the webui?

also whats your gpu?

[cursive cobalt]

These RPG Lora's use SDXL. All Lora's on the main drive (D) are visible, but none of the models from the secondary drive (E) are.

The Checkpoints are however visible, from both drives.

@ornate elk There :3 All info should be present in those screenshots

[ornate elk]

i dont see anything wrong there

[cursive cobalt]

Neither do i, but its not pulling from the secondary folders.

It does for the checkpoints, but not the loras

[ornate elk]

hmm

[cursive cobalt]

And its not a naming thing, cause it did regocnize this lora, that i just DL'd twice, 1 on each drive

Its also not an incompatibility thing, cause if it were, it wouldn't show either

[cursive cobalt]

I think i found the issue... Its exceeding the maximum number of dir "Path". I removed the "main" drive, and it showed E Drive

Re-added the main drive, and E vanished

[ornate elk]

ahhh

[cursive cobalt]

This however, is not optimal.

What can i do to bypass it?

Or circumvent the issues 4 pathing (without moving all to 1 drive) gives?

I gotta stop using big words, I struggle a bit with putting in a phrase correctly >.<

[ornate elk]

you can run the python installer again and there selecte "Disable Windows Path Lenght Limit"
but idk if it helps

[cursive cobalt]

That shouldn't in in of itself break anything, ye?

[ornate elk]

nope

i have it disabled too

[cursive cobalt]

Lets see if i can remember how to reinstall it

[ornate elk]

you only need to run the installer again

[cursive cobalt]

Thyere isn't any option to

[ornate elk]

at the end

[cursive cobalt]

Please stand by. I clicked a bit too fast, and ended up uninstalling it, so just gotta wait for that process to finish then reinstall xD

[cursive cobalt]

What end?

This is the launch window btw

[ornate elk]

Oh seems its only there when installing it and not modifying

[cursive cobalt]

Wasn't there when i was installing it. Looked everywhere, including under the installation path

Lemme just remove it again and reinstall, this time with pics of all steps xD

[ornate elk]

Install python 3.10.11

[cursive cobalt]

Wasn't that adviced against?

[ornate elk]

From who?

I recommend it in every guide

[cursive cobalt]

Seems to be a general concensus around installing Python for SD, but ok. We'll try 😄

[ornate elk]

Thats only because on github it says 3.10.6 is recommended. But it always works with a higher version to as long as its not 3.11 or higher

[cursive cobalt]

Ah

I c

[ornate elk]

I started using SD in 2022 with python 3.10.8

I'm on 3.10.11 since its release

[cursive cobalt]

Fair. I tried coding once. Spent like 3 hrs on a simply wizard. didn't work, so i gave up on coding xD

Also too large of a wall of text. My eyes blur stuff out

Add python to path?

Also gonna be doing a custom install to place in the same folder as previous version, so theres no change in anything that pulls python.

@ornate elk Nothing to change the limit

[ornate elk]

Maybe you already have it set

[cursive cobalt]

Extra question: How long does it usually take to restart/Reload the UI?

It seems to take an extraordinarily long time.

[proven vortex]

it seems to be working now but still no preview

[proven vortex]

3080

[random turtle]

Bit of an advanced question but maybe someone has done it before. I'm currently using Face Bounding Box with InsightFace to detect the face of a person and make an automatic crop. Upscale the automatic crop to 4k and than run a face fixer to add face detailers. After this is done I need to composite the cropped face back onto the original. I use compositedmasked and use the X/Y/width/height from the Face Bounding box to do this all automatically so it's always in the right place.

Now I'm also trying to do this for hands but I can't find a hand detector that also provides the X/Y/ coordiantes for comping it back on. Does anyone know a hand detector that can do this?

[azure sigil]

anyone here knows how often the api updates the remaining credits? I just noticed it's not real time

[mortal compass]

The fuck is this

[unreal knot]

Go to 127.0.0.1

I think

[mortal compass]

Yeah that is the one that usually launches. For some reason it just decided to go through localhost:7860 instead

Restarted it and it seems to repeat

[rough void]

@ornate elk it opens now but still the same error, could you send the tutorial agien?

[unreal knot]

Which UI do you use?

[mortal compass]

I switched from Forge to the standard A1111 today, which is probably why some things are going sideways

[mortal compass]

There is a search function top right

[unreal knot]

Did you keep your config file? It might be a mismatch if you didn't do a brand new directory

[mortal compass]

Yes I did, thought I might aswell so I don't need to redo all the settings, but didn't think that it would cause mismatches

I could just delete the current one, and it will download again, right

[unreal knot]

I'm just thinking since auto1111 had a brand new version recently and forge is now pretty old some of the files may be incompatible

[mortal compass]

Yeah that makes sense to me

I'll move the file somewhere else and see if that fixes anything

[neon kestrel]

anyone has an idea of what causes the gpu clock to be stuck at idle speed while trying to generate an image?

it seems like it goes up briefly when a image starts to cook and after a second or two it goes down to slightly above idle speeds

I've restarted auto1111 and it fixed the issue before, but all of sudden it comes back

[mortal compass]

Not really, but is that bad though? Are your generations affected by this by running slower or something in that manner

[neon kestrel]

oh yeah, you can see in the screenshot that it goes from 3~ s/iterations to 12ish

[rough void]

"Add the ´C:\ZLUDA´ folder and %HIP_PATH%bin to your PATH in the System Variables" idk what to do next

[mortal compass]

Oh I see. Well I'm just a novice, so can't help you there I think

[neon kestrel]

no worries

[random turtle]

awful screenshot coming it. But if there are two images coming out of a node in comfyui. How would I seperate them in the next node?

[mortal compass]

You're supposed to add the ZLUDA´ folder and %HIP_PATH%bin to your PATH, where the arrows are pointing

[rough void]

yeah did that, its this tab

[mortal compass]

then click new, and put one of the folder paths in, then do the same with the other

[rough void]

like that?

[mortal compass]

No, you need to insert the correct path to the wanted folder. Shift click on the wanted folder, then copy as path, then put the path inside the edit of your environment variables

[sinful needle]

do you know much about AnimateDiff

[rough void]

ok

[mortal compass]

I don't I'm afraid, never used it

[random turtle]

If I draw two previews it's still two previews with two images. I assume there is a way to split it up?

[random turtle]

Do you have any idea how to split up the images?

seems like a really simple thing googled everywhere but didnt find an answer

[mortal compass]

Menar inte alls att låta nedlåtande, jag var själv en färsking för bara 2 månader sedan; hur mycket kan du om programmering och kod?

[rough void]

0

[rough void]

menar du denna?

[mortal compass]

Ait samma som mig då. För att kunna klara dig igenom den riktigt svåra biten, alltså den kurvan i början av när man lär sig något helt nytt, så måste du nästan studera hur A1111 och Stable Diffusion funkar. Kolla upp videos med tutorials, hur man installerar, läs poster på forum osv osv. Annars kommer du inte komma långt, enligt min egna erfarenhet iallafall

[mortal compass]

Yes, högerklick med shift på den

Sen matar du in den i environment variables menyn

[rough void]

ska jag dra ut den o lägga den på c disken?

[mortal compass]

Alltid bra att ha sina grejer rakt upp och ned på C: disken

Tänk på att inte ha mellanslag i dina filnamn/mappnamn också

@rough void Have you checked out posts on reddit or videos about how you install the zuda mod? If not, I strongly recommend doing so

[rough void]

i dont get the error now but still no img

[mortal compass]

I don't know what you mean by "still no image"

[rough void]

and after its done loading "nothing"

[mortal compass]

When you're troubleshooting it is always very valuable if you can be as precise as possible by adding context, details, etc. Like, I can't see whats wrong only by a picture of the UI

Can you show the terminal

[rough void]

idk what that is

[mortal compass]

cmd

[rough void]

yeah?

is that not it?

[mortal compass]

Yes this is correct

I think this is too advanced for me unfortunately. I'm still a noob regarding all this, and I don't wanna tell you to do stuff that might not work or that will screw something up you know

[rough void]

ok this is the first page btw

[mortal compass]

Bumping this for you, this would help you a lot, although it takes a bit of time.

[random turtle]

I don't want to pressure you since you also just doing this in your own time but it seems you know your way around comfy. Any chance you can quickly look at my question. It seems very simple but I just can't find anyway to seperate it

[mortal compass]

Oh, I think there's a lot of things that has gone wrong with your installation

[mortal compass]

I've actually never used comfy, it scares the hell out of me

[rough void]

@ornate elk we need you

[mortal compass]

Never bothered with learning it, I got no programming or coding history so to learn comfy would just take way too much time

[jagged glade]

@ornate elk bro issue not solved please provide a video

i called my frnd to screen sharing and still cant fix it

[mortal compass]

Guys chill with the pings, he seems to not be available at the moment

[jagged glade]

AsyncConnectionPool.init() got an unexpected keyword argument 'socket_options'

ok anyone else wanna help me in this

[mortal compass]

Have you googled AsyncConnectionPool.init() or other variations of it

[jagged glade]

tried it

[random turtle]

Ah thanks anyway!

[jagged glade]

venv is the issue how to reset pip and library?

[mortal compass]

Alright, well, I don't know what could be the problem. I'd try asking the new chatgpt4o, it's really impressive

If you think venv is the issue, you can always delete it

It will redownload upon next boot of the program

[jagged glade]

leaving it here for anyone else

[mortal compass]

I'm not telling you to do this, but many times when I've had errors or gotten stuck with something that won't get better, I've just deleted everything and started over the installation process

I'm actually gonna do it right now lol

[jagged glade]

last 1 week

already reinstalled webui but i have maybe delete everything

[jagged glade]

done already same issue

[mortal compass]

Then I am out of ideas, sorry lads

[jagged glade]

u can ask any admins?

[mortal compass]

I'm just a pleb like you, I have no power here

Several people have pinged CS1 so he will be here when he sees it

Can't do much more

[lean badger]

Just now I subscribed on stability.ai website for image generation. But unable to generate images here, what's the drill guys?

[rough void]

i had some issues with my wifi when i installed it, its good now. do you think it will help if i try and reinstall it

[mortal compass]

I would try installing it, yes. Try to read every step several times, so you don't miss anything. It's easy to do so and then end up with nothing working

[terse lichen]

what does mean tho ?

[mortal compass]

No idea, sorry!

[mortal compass]

Looks like your cpu cant handle the amount of allocated memory

[rough void]

space or no space?

[mortal compass]

always a space before --commands

[rough void]

ok

[rough void]

do i need to reinstall zluda aswell?

[mortal compass]

Wait, do you have a working SD web UI atm

[rough void]

?

[mortal compass]

As far as I understand, Zuda is a mod for stable diffusion

you cant install a mod without having the game, so to speak

[ornate elk]

@jagged glade what was your error again?

[unreal knot]

Does anyone here have experience with locally run SD discord bots? I can't for the life of me find one that can handle LORAs

[jagged glade]

This one

[rough void]

but i have the game

[jagged glade]

The computer power cutoff and after that day this error show up

[ornate elk]

It says None for the version and also dreambooth always breaks the webui

So a proper reinstall would be the best

[jagged glade]

How to uninstall all the softwares? Bcz I download new webui and error is same

[mortal compass]

Okay so you're saying, you have all the correct versions of Python, Git, and you have installed the programs so they are in the PATH directory, correct?

[rough void]

yeah but i need to and %hip thing, do you know how to do that

[mortal compass]

I'm not familiar with that I'm afraid

[rough void]

https://forum.eset.com/topic/15740-environment-variables-for-hips-rules/

i dont get it

[mortal compass]

Yeah that kinda computer language is too advanced for me

[rough void]

i think i did it

but now i got this

[mortal compass]

I don't recognize this at all.. Sorry dude

[open fox]

It seems either the connection was cut off or mutiple sessions opened or whatever. In either case the base file which should be downloaded was not downloaded completly.
After 1.1 Gbyte from 2.7 Total it broke the connection. You Could simply retry

[rough void]

i did it agien now i got this

[rough void]

lord before you type

[open fox]

Same error but later in the process.... Restart again, as packages already installed will be skipped

[rough void]

i did somthing dif

[rough void]

i dident have "hip path" vefore did i add it right?

[ornate elk]

Follow my install guide in the pinned messages of this channel
You can move your models over if it works

[jagged glade]

Ok

[open fox]

@ornate elk great timing as nvidia owner the complete HIP / ROCM ZLUDA Stuff is far far away...

[open fox]

So with the thumbs up from CS1o: yes it is correct 🙂

[jagged glade]

delete the whole folder?

[viral knot]

delete system32

[rough void]

dose this look right?

[open fox]

At least no error appears. Just try it

[rough void]

dont work

just loading

[open fox]

So you open http://127.0.0.1:7860 an it does not load

[rough void]

it dose

but not the img

[open fox]

What does the Command Window shows?

[rough void]

wdym

[open fox]

The screenshot from 20:27 was taken while you already tried to generate a image?

[rough void]

yes

before

[open fox]

No i meant the black screensoht

of the command line

[viral knot]

white screen, my eyes!

[open fox]

Please retry to generate an image and then take another screenshot of the command window.

[rough void]

before after

and it stays at 0%

[open fox]

What does the taksmanager says. Usage of Memory, CPU and GPU?

[rough void]

oh

what the flip

[open fox]

Well a face.... not happy but still a face

[rough void]

how do i make it "not suck"

[viral knot]

just add smiling face 🙂

[open fox]

And try duck face for more sucking 😄

[viral knot]

i saw that in my nightmares

[rough void]

i'm scared

[viral knot]

wait what model are you using?

[rough void]

it needed a trigger word

[open fox]

Well you are now technically enabled to generate pictures. Now the learning progress starts 🙂

[viral knot]

eventually he will learn comfy :3

[viral knot]

most loras need triggers, but some dont, you have to check the lora model page for details

[rough void]

do you use , or pace when starting a new thing?

or enter

[open fox]

Depends on model architeture and comfyui workflow i use, most t5 base can understand sentences even without weights etc. The general one work fine with comma separated lists and weighting.

[rough void]

ok ty for help

[rough void]

anyway to keep the face details and quality of the face even when doing figure drawing?

[jagged glade]

downloading all models now

[bright light]

just use inpainting to fix faces

[cosmic furnace]

@ornate elk Dyu know how fast amd on zluda now compared to nvidia and to how it is on linux?)

[ornate elk]

some guy made a webui himself using zluda and it gets 43 it/s on 1.5 models and 10 it/s on sdxl

with an 7900xtx

[cosmic furnace]

what dya mean made webui himself o__o

[ornate elk]

he coded one

[cosmic furnace]

its just... ive heard on linux amd is still pretty much faster than its now on zluda y ? =\

[ornate elk]

thats true

[cosmic furnace]

ive seen a video . and yeah maybe thats not some fair & squre clean test. And yes super version is an improve of usual. R there gonna be 4060-4090 super's btw?

[ornate elk]

i dont think there will be any new super cards

the rtx 50 series is coming

[cosmic furnace]

But for me tests shown that 4070 super with 12 gb on a youtube video is an 7-10 sec faster ? o.0 . Just same as shim mostly juggernaut model v8. sdxl . 100 steps lizard . same euler A . etc. At least what i can . promt always same , maybe lil bit shorter

While ive heard on linux even 7900 xt is about as fast as 4080 or even lil bit faster

which means yes, its not clean test but still 😦 . Zluda is not yet at it full now ?

[ornate elk]

yea zluda support is not fully there

[cosmic furnace]

btw for some reasons the steps-promt and that model... Well repeating his settings ended up in it\s almost twice lower than usual

maybe when alot of steps dunno ?

[ornate elk]

idk, maybe some wrong setting

[cosmic furnace]

how do you make yourself this webui xDD. wait ive heard ppl dont like webui one? Auto kinda pretty much more popular y ?

[ornate elk]

i dont get that question xD

[cosmic furnace]

DYu know how to zluda webui? And ppl mostly prefer Auto over webui for a reason y ?

[hallow lance]

Hello, could anyone ask me a question, how do you make the images? I'm having a really hard time finding apps to do.

[ornate elk]

Auto is a webui

[cosmic furnace]

Oh) i mistook it for comfy >_<

[sinful needle]

[serene laurel]

Hello. I have a problem using stablediffusion API. Could I write my question here, or is there another place?

[velvet bridge]

what is your problem

[dense tapir]

Hello just wanted to say that @ornate elk is a goat

[wet flame]

i get similar errors after few image generation too

idk if its something to do wth my older ram and mb

i use ddr4

[terse lichen]

is it cause of the high upscale limit ?

[dense tapir]

Would someone please help me with this?

no module 'xformers'. Processing without...

I have no clue why I get this

[quasi ginkgo]

Do you have the xformers line in your webui user .bat?

[dense tapir]

Yes

[quasi ginkgo]

AFAIK Xformers isnt a critical component, just helps with optimization

[dense tapir]

Is it true that it makes iterations come out faster?

[quasi ginkgo]

So people say, I haven't noticed a difference tho

Anyone else having trouble with "NansException: A tensor with all NaNs was produced in Unet."? I did a discord search and saw a few other people were having issues with it but couldn't find a definitive fix. Seen people say it might be ADetailer resource usage related. File "C:\SD\stable-diffusion-webui\modules\call_queue.py", line 57, in f res = list(func(*args, **kwargs)) File "C:\SD\stable-diffusion-webui\modules\call_queue.py", line 36, in f res = func(*args, **kwargs) File "C:\SD\stable-diffusion-webui\modules\txt2img.py", line 109, in txt2img processed = processing.process_images(p) File "C:\SD\stable-diffusion-webui\modules\processing.py", line 845, in process_images res = process_images_inner(p) File "C:\SD\stable-diffusion-webui\extensions\sd-webui-controlnet\scripts\batch_hijack.py", line 59, in processing_process_images_hijack return getattr(processing, '__controlnet_original_process_images_inner')(p, *args, **kwargs) File "C:\SD\stable-diffusion-webui\modules\processing.py", line 981, in process_images_inner samples_ddim = p.sample(conditioning=p.c, unconditional_conditioning=p.uc, seeds=p.seeds, subseeds=p.subseeds, subseed_strength=p.subseed_strength, prompts=p.prompts) File "C:\SD\stable-diffusion-webui\modules\processing.py", line 1328, in sample samples = self.sampler.sample(self, x, conditioning, unconditional_conditioning, image_conditioning=self.txt2img_image_conditioning(x)) File "C:\SD\stable-diffusion-webui\modules\sd_samplers_kdiffusion.py", line 218, in sample samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, x, extra_args=self.sampler_extra_args, disable=False, callback=self.callback_state, **extra_params_kwargs)) File "C:\SD\stable-diffusion-webui\modules\sd_samplers_common.py", line 272, in launch_sampling return func() File "C:\SD\stable-diffusion-webui\modules\sd_samplers_kdiffusion.py", line 218, in <lambda> samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, x, extra_args=self.sampler_extra_args, disable=False, callback=self.callback_state, **extra_params_kwargs)) File "C:\SD\stable-diffusion-webui\venv\lib\site-packages\torch\utils\_contextlib.py", line 115, in decorate_context return func(*args, **kwargs) File "C:\SD\stable-diffusion-webui\repositories\k-diffusion\k_diffusion\sampling.py", line 145, in sample_euler_ancestral denoised = model(x, sigmas[i] * s_in, **extra_args) File "C:\SD\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1518, in _wrapped_call_impl return self._call_impl(*args, **kwargs) File "C:\SD\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1527, in _call_impl return forward_call(*args, **kwargs) File "C:\SD\stable-diffusion-webui\modules\sd_samplers_cfg_denoiser.py", line 269, in forward devices.test_for_nans(x_out, "unet") File "C:\SD\stable-diffusion-webui\modules\devices.py", line 255, in test_for_nans raise NansException(message) modules.devices.NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.
Heres some hopefully relevant logs. All I can decipher out of it is that devices.py is doing a nans check which is coming out truthy. I got a 12gb vram 3080. I've found that I can get passed it by changing models back and forth or by adding/removing loras, but its still a pain to deal with

[terse lichen]

hmm

[open fox]

Guess you are out of memory and it swaps. So it get's real slow here.

[vague bronze]

hello

[terse lichen]

so how can i check which is being eaten a lot ?

[open fox]

Try a lower resolution for the generation / latents. AnimateDiff is kind of hungry. You could also try to reduce context length.

[terse lichen]

which is the upscale which needs to be lowered

[open fox]

I don’t know the workflow you are using. I only can assume that the workflow requires a lot of memory by the length one iteration takes. So reducing the footprint by changing the resolution (height x width) of input latents (or source videos,…) should help.

[ornate elk]

Yes, and it reduces the vram usage

Make sure you always launch the webui-user.bat and not the webui.bat

[ornate elk]

What is your txt2img setting?

[quasi ginkgo]

Euler A, 35 sample steps, 768x768, batch size 1, batch count 1, no hires fix or refiner, ADetailer enabled

[ornate elk]

It seems like you have Controlnet enabled too?

[quasi ginkgo]

Negative, I checked and the "enabled" box isn't checked

[ornate elk]

Strange. Because the error mentioned Controlnet

What model do you use?

[quasi ginkgo]

autismmixSDXL_autismmixPony

[ornate elk]

Does it work when Adetailer is disabled?

[quasi ginkgo]

Hit or miss, works more consistently with it disabled but still happens quite frequently

[ornate elk]

You have a 3080 right?

[quasi ginkgo]

Yessir

[ornate elk]

Okay, then make sure you have --xformers --medvram-sdxl --no-half-vae
In your webui-user.bat
And then try again

[quasi ginkgo]

10-4, I'll add it - run it a few times - and see if that changes anything

huh
guess xformers wasnt installed idk if that'll help but it sure dont hurt to have it now

[ornate elk]

It should be much faster now

[quasi ginkgo]

Got hit with it again

      File "C:\SD\stable-diffusion-webui\modules\call_queue.py", line 57, in f
        res = list(func(*args, **kwargs))
      File "C:\SD\stable-diffusion-webui\modules\call_queue.py", line 36, in f
        res = func(*args, **kwargs)
      File "C:\SD\stable-diffusion-webui\modules\txt2img.py", line 109, in txt2img
        processed = processing.process_images(p)
      File "C:\SD\stable-diffusion-webui\modules\processing.py", line 845, in process_images
        res = process_images_inner(p)
      File "C:\SD\stable-diffusion-webui\extensions\sd-webui-controlnet\scripts\batch_hijack.py", line 59, in processing_process_images_hijack
        return getattr(processing, '__controlnet_original_process_images_inner')(p, *args, **kwargs)
      File "C:\SD\stable-diffusion-webui\modules\processing.py", line 981, in process_images_inner
        samples_ddim = p.sample(conditioning=p.c, unconditional_conditioning=p.uc, seeds=p.seeds, subseeds=p.subseeds, subseed_strength=p.subseed_strength, prompts=p.prompts)
      File "C:\SD\stable-diffusion-webui\modules\processing.py", line 1328, in sample
        samples = self.sampler.sample(self, x, conditioning, unconditional_conditioning, image_conditioning=self.txt2img_image_conditioning(x))
      File "C:\SD\stable-diffusion-webui\modules\sd_samplers_kdiffusion.py", line 218, in sample
        samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, x, extra_args=self.sampler_extra_args, disable=False, callback=self.callback_state, **extra_params_kwargs))
      File "C:\SD\stable-diffusion-webui\modules\sd_samplers_common.py", line 272, in launch_sampling
        return func()
      File "C:\SD\stable-diffusion-webui\modules\sd_samplers_kdiffusion.py", line 218, in <lambda>
        samples = self.launch_sampling(steps, lambda: self.func(self.model_wrap_cfg, x, extra_args=self.sampler_extra_args, disable=False, callback=self.callback_state, **extra_params_kwargs))
      File "C:\SD\stable-diffusion-webui\venv\lib\site-packages\torch\utils\_contextlib.py", line 115, in decorate_context
        return func(*args, **kwargs)
      File "C:\SD\stable-diffusion-webui\repositories\k-diffusion\k_diffusion\sampling.py", line 145, in sample_euler_ancestral
        denoised = model(x, sigmas[i] * s_in, **extra_args)
      File "C:\SD\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1518, in _wrapped_call_impl
        return self._call_impl(*args, **kwargs)
      File "C:\SD\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1527, in _call_impl
        return forward_call(*args, **kwargs)
      File "C:\SD\stable-diffusion-webui\modules\sd_samplers_cfg_denoiser.py", line 269, in forward
        devices.test_for_nans(x_out, "unet")
      File "C:\SD\stable-diffusion-webui\modules\devices.py", line 255, in test_for_nans
        raise NansException(message)
    modules.devices.NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.

idk if thats any different
ADetailer re-enabled on restart because I have it preset that way, Controlnet is still disabled

[mortal compass]

Might be, yes. It does the same on my pc. What I do is that I generate images first, then I upscale the ones I like to x2, x3 if I'm feeling cheeky. x4 is kinda redundant if you ask me. What's your upscale workflow?

[terse lichen]

this is the model i am using @mortal compass its ipiv animatediff workflow

[mortal compass]

Ooh jeez that's impressive

But I don't understand anything of it lol

[terse lichen]

its quite a big workflow or heavy is the right term @mortal compass

[mortal compass]

Yeah I can see that! I hope you get it to work since you spent a lot of time creating this workflow

[terse lichen]

its by another person

[open fox]

Maybe try installing the Crystools nodes with the manager. It shows the ressourcen within comfy. So you can see how your vram is filled and which node leads to the 100% usage

[ornate elk]

Any lora used that is maybe not compatible with pony?

[terse lichen]

this is just what i was looking for bro

so how to read this information bro @open fox

the bar is fluctuating between cpu and gpu

[terse lichen]

so in short animatediff is a hungry model and needs a lot of gpu power and memory to run @open fox @mortal compass

[mortal compass]

All I know is that some models can be really hungry for VRAM. Like SUPIR upscaling reacquires a lot of power, and I also know that animatediff is one of those that needs more. Is it possible to look up the requirements perhaps?

[terse lichen]

it needs 10gb vram now and before it was 60 gb vram requirement .....

[mortal compass]

lmao 60

And u have atm?

[terse lichen]

a 3060 card

they have lowered it now to 10gb

[mortal compass]

surly you know how much vram your card has

[terse lichen]

it is a 12 gb card

[mortal compass]

I see. Well, have you made a post at comfy's github page asking for help? Also there are really good articles regarding everything AI generated on the civitai website, might be something there which you can use

[unreal stone]

What steps can we take to ensure that the images generated by Stable Diffusion resemble realistic human appearances?

[mortal compass]

That's a very deep and sometimes complicated subject, but I suggest you go here https://civitai.com/articles and take a look around. Lots of good articles made by users.

The TLDR is basically you combine (either one or several, depends) checkpoints, loras and embeddings to avoid txt2img abominations. Good prompting is also key to this.

[open fox]

Well you can be sure that if the bar changes from CPU anf Memory to GPU and VRAM and that constantly, or like in the first screenshot where CPU and Memory is high the calculations are done by your CPU not GPU. This is significant slower. So you get numbers like 1600 seconds for one iteration.

[terse lichen]

i see

right now its this

[open fox]

Yeah but you see that your memory is still used. So either it does things simultanios or it swaps between RAM and VRAM and this takes time. Sadly. Is it the basic morph workflow from ipvi or another one?

[terse lichen]

its from ipvi

[open fox]

https://civitai.com/models/372584?modelVersionId=469548# This one

?

[terse lichen]

yes

[open fox]

The comments say that a 12gbyte card should work.

[terse lichen]

its a 3060 card with 12gb vram

[unreal stone]

I see, thanks for your prompt response Also, it keeps blurring SFW images. Mostly, if the character is a female. But happens the other way too

[terse lichen]

in short you need to adjust the prompt and also the steps are mainly for how deatiled you want the image at the cost of generation time and your vram

[mortal compass]

Negative prompts is your best friend, you can put anything in there. For example, if you get blurry images, you can try something like:
prompt:sharp focus, 8k, UHD, best quality
neg prompt: blurry, bad focus, bad quality

Prompting is also very dependent on which models you're using. To get good prompt examples, go to that civitai and search for something related to your image idea. Like "photo of a womans profile in the sunlight" or whatever you want. If you click on the images in your result, you can see their prompts, models used, etc etc

[solemn moon]

who updated control net, and why??? i lost my sweet spot, what changed? who is yr manager

jk obvsies but theres sum truth to it innit

[open fox]

just tried the workflow.

Just used some dummy reference images

[terse lichen]

its a good work just trying to understand it or make it work

[open fox]

Max Memory taken was 17 Gbyte. But i think it is because i increased the base resolution. Otherwise the workflow worked out of the box

[terse lichen]

thats i am trying to see if it can be used in my work

[open fox]

so does it work now with your setup? Is it possible to generate the video?

[terse lichen]

its working but its heavy load on the cpu gpu and memory

[vernal hare]

hi all

i'm trying to create variations of an image using the api

however, all of the variations I generate are identical

does anyone have any idea why this might be?

[terse lichen]

if you want to have variations in your generation you need to tweek your prompts

[unreal stone]

I've tried using these prompts, but the images, especially those containing females, are still mostly blurred. Is there a method to detect if an image has been blurred or flagged, so that I can take appropriate action? It just returns the image bytes and does not tell if it is blurred or not

[mortal compass]

What checkpoint are you using?

[vernal hare]

Fixed this by tuning the schedule parameters

Thanks anyways guys

[unreal stone]

I am using Stable Diffusion 3

[mortal compass]

SD 3 is not a checkpoint

[unreal stone]

I am new to this, can you help me out? I am not sure what's a checkpoint

[mortal compass]

Yes I am trying my man 🤣

A checkpoint is a model that has been trained through machine learning. If you want to generate pictures, first, you need a program. You picked SD 3's API which I know nothing about, so my knowledge is very limited regarding it. Second, you need a checkpoint (ckpt). There are a huge amount of checkpoints for SDXL, but I have no clue about SD 3

Anyway, these models are often made for specific things. Such as realism, NSFW, anime, animals, landscapes or whatever, you name it

So you need to pick a model that fits what you want to generate

[mortal compass]

If you have an interest for txt2img I highly suggest that you go to youtube, and watch A TON of videos, or just read plain articles about it if videos are not your thing. It's a very deep and complicated subject

[unreal stone]

Thanks for your guidance. I will definitely do these.

[mortal compass]

Anytime! Be patient, I was also 100% new to this 2 months ago. Now I feel much more comfortable with all of this! Ofc if you get stuck you are welcome to write here:)

[unreal stone]

My main purpose is to generate realistic images, for that I wanted to use midjourney but it doesn't have a public api. DALL-E produces very unrealistic cartoonish images. So, I tried Stability's SD3 API. Do you have any suggestions or a path for me

[mortal compass]

I'm using Stable Diffusion locally, I don't know much about API's. I know that https://civitai.com/ has a very good online generator that I'm using daily, the catch is that you need their currency to make them. You can get currency by doing daily "quests" and if you get reactions on your images, no need to oay for it if you don't want to.

[jagged glade]

not a tech question but which one is SD 1.9.3?

i used 1.5 until last week i wanna know is it SD 2?

(talking about the base model in filters)

[violet swift]

I installed Stylegan, but when I click on Install Missing Custom Nodes or Install Custom Nodes, the missing nodes do not appear

[onyx solar]

Stable Diffusion model versions are listed within that Base Model section.

1.9.3 is not a version of Stable Diffusion, it's a version of Automatic1111, one of the various apps to generate images with SD models.

[lost crescent]

Reactor Nodes and WAS Nodes - just how do you get them to work inside ComfyUI?!?!?!?

[onyx solar]

Same way you do any nodes? Install the custom nodes, add to workflow, drag and connect dots, generate image, marvel at outcome.

[lost crescent]

Yes - but always in RED - and says during startup - Cannot Import Was/REACTOR Nodes 😦

[onyx solar]

And do you have ComfyUI Manager installed? Have you tried the Install Missing Nodes function? If so, when was the last time you updated ComfyUI?

[cinder ermine]

im trying to train a style lora, the style itself works quite well, but details such as faces are lost. Should i add cropped pictures of just the faces to the dataset?

[violet swift]

help plsss aaaa

[sinful needle]

anyone here know anything about animatediff?

[open fox]

Maybe you could precise your question a bit. I use it sometimes but that does not mean I can answer all questions

[vocal burrow]

Sd 1.9.3 does not exist you re probably confusing your webui/software version with actual sd model version (1, 1.5, 2, xl, 3, etc)

[ornate elk]

Yes

[sinful needle]

Well I've been trying to get your attention, sir wizard, but to no avail. 😦

This only happens when I use LORAs with animatediff it appears.

[ornate elk]

try set the batch size of animatediff to 4 and try again

[copper fulcrum]

Ive set my Clip Skip setting to 2 but whenever I generate an image and reload the settings from it - it says "Clip skip 1" in the override box... how do I fix this?

[crude halo]

Im using XYZ script to generate multiple variants of clip skip with diff seeds. Problem is, 1)it inverts clip skip and seeds. 2)it only returns 1 seed per clip skip. Am I doing something wrong?

[ornate elk]

Looks correct. Clip skip 1 has 5 images,
clip skip 2 aswell and so on

[crude halo]

yes, but the different seeds produce seemingly identical images

also rows and columns are inverted

[ornate elk]

you get 5 different images each time,
clip skip doesnt change much or anything in sdxl so thats why they look the same with different clip skip

[wind thunder]

my model dont load ?

maybe i write it wrong ?
set COMMANDLINE_ARGS=--use-zluda --lowvram --update-check --skip-ort

[quasi ginkgo]

I have noncompatible loras hidden

[wind thunder]

wait @ornate elk can i run ZLU SD with a NVIDIA geforce GTX 1060 6GB ?

[ornate elk]

no, zluda is for AMD

[crude halo]

you're right, I made a mistake with other testings I made

[proper pike]

Error(s) in loading state_dict for Resampler:
size mismatch for proj_in.weight: copying a param with shape torch.Size([1280, 1280]) from checkpoint, the shape in current model is torch.Size([1280, 1664]).

File "E:\ComfyUI_windows_portable\ComfyUI\execution.py", line 151, in recursive_execute
output_data, output_ui = get_output_data(obj, input_data_all)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\ComfyUI_windows_portable\ComfyUI\execution.py", line 81, in get_output_data
return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-0246\utils.py", line 381, in new_func
res_value = old_func(*final_args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\ComfyUI_windows_portable\ComfyUI\execution.py", line 74, in map_node_over_list
results.append(getattr(obj, func)(**slice_dict(input_data_all, i)))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI_IPAdapter_plus\IPAdapterPlus.py", line 711, in apply_ipadapter
work_model, face_image = ipadapter_execute(work_model, ipadapter_model, clip_vision, **ipa_args)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "E:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI_IPAdapter_plus\IPAdapterPlus.py", line 340, in ipadapter_execute
ipa = IPAdapter(
^^^^^^^^^^
File "E:\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI_IPAdapter_plus\IPAdapterPlus.py", line 70, in __init__
self.image_proj_model.load_state_dict(ipadapter_model["image_proj"])
File "E:\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\nn\modules\module.py", line 2189, in load_state_dict
raise RuntimeError('Error(s) in loading state_dict for {}:\n\t{}'.format(``` does anyone know the fix?

[covert lotus]

are lycoris models able to be used on the amd 1111 version of SD? whenver i try one it crashes

D:\a_work\1\s\pytorch-directml-plugin\torch_directml\csrc\engine\dml_util.cc:118] Invalid or unsupported data type.

[dense rune]

hello, someone knows how to avoid these lines when upscaling or how to fix it after upscaling? this is a 8196x12288 image

[ornate elk]

Yes they should work with directml and zluda

[dense rune]

tried with tile resample with controlnet is more important

[ornate elk]

Lower the denois below 0.3

[dense rune]

i use denoise 0.2

[ornate elk]

Ohh

Do you use sdxl or 1.5?

[dense rune]

1.5

[covert lotus]

hm thats odd any reason why i might get that error message then?

its only when i try using a lycoris file

[ornate elk]

would need to see the full cmd log also, whats your gpu?

[covert lotus]

rx 6800

[covert lotus]

just paste the whole thing?

[ornate elk]

yep

[covert lotus]

alright lemme make it crash rq

thx for the help btw

https://civitai.com/models/275291/sdxl-artist-style-mika-pikazo?modelVersionId=525768 this is what im tryna use

venv "C:\Users\sharr\Desktop\SD\stable-diffusion-webui-amdgpu\venv\Scripts\Python.exe"
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Version: v1.9.3-amd-13-g517aaaff
Commit hash: 517aaaff2bb1a512057d88b0284193b9f23c0b47
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
C:\Users\sharr\Desktop\SD\stable-diffusion-webui-amdgpu\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: --autolaunch --theme dark --skip-version-check --disable-nan-chec --use-directml --upcast-sampling --precision full --lowvram
ONNX: version=1.18.0 provider=DmlExecutionProvider, available=['DmlExecutionProvider', 'CPUExecutionProvider']
Loading weights [e3c47aedb0] from C:\Users\sharr\Desktop\SD\stable-diffusion-webui-amdgpu\models\Stable-diffusion\animagineXLV31_v31.safetensors
[LyCORIS]-WARNING: LyCORIS legacy extension is now loaded, if you don't expext to see this message, please disable this extension.
Running on local URL: http://127.0.0.1:7860

To create a public link, set share=True in launch().
Creating model from config: C:\Users\sharr\Desktop\SD\stable-diffusion-webui-amdgpu\repositories\generative-models\configs\inference\sd_xl_base.yaml
Startup time: 13.1s (prepare environment: 16.2s, initialize shared: 2.0s, load scripts: 1.6s, create ui: 0.6s, gradio launch: 0.5s).
C:\Users\sharr\Desktop\SD\stable-diffusion-webui-amdgpu\venv\lib\site-packages\huggingface_hub\file_download.py:1132: FutureWarning: resume_download is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use force_download=True.
warnings.warn(
Applying attention optimization: InvokeAI... done.
Model loaded in 24.0s (load weights from disk: 1.0s, create model: 0.6s, apply weights to model: 19.5s, apply half(): 0.2s, hijack: 0.1s, calculate empty prompt: 2.3s).
0%| | 0/20 [00:00<?, ?it/s][F D:\a_work\1\s\pytorch-directml-plugin\torch_directml\csrc\engine\dml_util.cc:118] Invalid or unsupported data type.

[covert lotus]

hopefully this is what u needed :)

[ornate elk]

oh wow there is a lot wrong in your webui-user.bat
first remove --skip-version-check --disable-nan-chec --precision full --lowvram
and add: --medvram --no-half-vae --opt-sub-quad-attention --opt-split-attention

then relaunch the webui-user.bat and try again

and then show me your Extensions

you might have the old Lycoris extension still installed

[covert lotus]

ok so i did the first fix and it had the same error

[ornate elk]

But in any way i dont recommend to use the AMD Directml webui at all anymore. Your card can use the Webui with Zluda support. which is much faster and let you generate stuff without going out of vram

[ornate elk]

ah well you have the lycoris extension installed, its not needed as lycoris work the same as loras now, you can delete it, and put the Lycoris into the models/lora folder, and then it should work

@covert lotus If you want to try the faster Webui Version for AMD, you can follow my install Guide from the Pinned Messages of this Channel, or directly linked here:
#🤝｜tech-support message

[covert lotus]

i'll deffo look into it, was scratching my head earlier

[ornate elk]

im on AMD too, if you need help with any install Step

[covert lotus]

i did this btw and its still not working

[ornate elk]

new error in the cmd?

[covert lotus]

although i think i might've just disabled the lycoris extension not delete it, will that make a diff?

same one

[ornate elk]

you need to delete it and restart

[covert lotus]

how exactly do u delete it not just deactive it

[ornate elk]

you have to delete it from the extensions folder

[covert lotus]

ah ok lemme try now

[covert lotus]

and thanks will deffo try ur guide

[covert lotus]

same error

[ornate elk]

can you show the log?

[covert lotus]

not quite sure how sorry...

im not very good at this lmao

[ornate elk]

i mean the cmd text again

[covert lotus]

venv "C:\Users\sharr\Desktop\SD\stable-diffusion-webui-amdgpu\venv\Scripts\Python.exe"
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Version: v1.9.3-amd-13-g517aaaff
Commit hash: 517aaaff2bb1a512057d88b0284193b9f23c0b47
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
C:\Users\sharr\Desktop\SD\stable-diffusion-webui-amdgpu\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: --autolaunch --theme dark --use-directml --upcast-sampling --medvram --no-half-vae --opt-sub-quad-attention --opt-split-attention
ONNX: version=1.18.0 provider=DmlExecutionProvider, available=['DmlExecutionProvider', 'CPUExecutionProvider']

You are running torch 2.0.0+cpu.
The program is tested to work with torch 2.1.2.
To reinstall the desired version, run with commandline flag --reinstall-torch.
Beware that this will cause a lot of large files to be downloaded, as well as
there are reports of issues with training tab on the latest version.

Use --skip-version-check commandline argument to disable this check.

Loading weights [e3c47aedb0] from C:\Users\sharr\Desktop\SD\stable-diffusion-webui-amdgpu\models\Stable-diffusion\animagineXLV31_v31.safetensors
Running on local URL: http://127.0.0.1:7860

To create a public link, set share=True in launch().
Creating model from config: C:\Users\sharr\Desktop\SD\stable-diffusion-webui-amdgpu\repositories\generative-models\configs\inference\sd_xl_base.yaml
Startup time: 13.7s (prepare environment: 17.7s, initialize shared: 2.2s, load scripts: 1.3s, create ui: 0.6s, gradio launch: 0.5s).
C:\Users\sharr\Desktop\SD\stable-diffusion-webui-amdgpu\venv\lib\site-packages\huggingface_hub\file_download.py:1132: FutureWarning: resume_download is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use force_download=True.
warnings.warn(
Applying attention optimization: Doggettx... done.
Model loaded in 24.9s (load weights from disk: 1.0s, create model: 0.5s, apply weights to model: 20.2s, apply half(): 0.1s, apply dtype to VAE: 0.1s, move model to device: 0.1s, hijack: 0.1s, calculate empty prompt: 2.6s).
0%| | 0/28 [00:00<?, ?it/s][F D:\a_work\1\s\pytorch-directml-plugin\torch_directml\csrc\engine\dml_util.cc:118] Invalid or unsupported data type.
Press any key to continue . . .

could it be the 2.0.0 torch ver?

[ornate elk]

nope

i test the lora real quick on my directml version

[covert lotus]

And it was fine I take it

I’m not sure at all what it could be lmao

[ornate elk]

ok i get the same error

im gonna try it now on the webui with zluda

[covert lotus]

Alright fingers crossed

[ornate elk]

it works

no error there

[covert lotus]

So essentially delete stable diffusion that I have installed and restart using ur guide?

[ornate elk]

first install the one using my guide
then if everything works, you can move your models and loras over and then delete the other one

[covert lotus]

My install is fresh from today literally nothing else on it

So I don’t need anything

[ornate elk]

also move on model from the old install to the new one to speed the installation up

[ornate elk]

oh then you can delete it sure

[covert lotus]

Is the guide process long? Or around the same as a normal set up

[ornate elk]

the first part is mostly the same and can be done in under 5 minutes

the zluda part can be tricky, but nothing special (10 minutes)

and the first image generation will take 15-40minutes (only one time)

[covert lotus]

Sweet ty my friend

[ornate elk]

here is mika pikazo style ^^ idk if its accurate

[covert lotus]

I’m going to sleep I’ll message you tomorrow if it all works

[covert lotus]

Small forehead aside yeah that’s great

[ornate elk]

okay, sure, let me know if you have any questions

[ornate elk]

Can you show me the img2img settings you used?
I can take a look at it tomorrow

[dense rune]

thanks

[distant topaz]

Anybody know why I'm getting these Jackson Pollock paintings? Still new to SD so no big words please xD

[tidal steeple]

I'm confused. I copied a workflow directly from a civitai picture someone produced, but every time I try to run the same exact workflow, I get outputs like this... Anyone know what's going on?

[onyx solar]

You have the wrong VAE chosen.

[tidal steeple]

This workflow was pulled straight from civitai. Is it an issue with the VAE being old or something? I've got the same one they used

[onyx solar]

You are using an SDXL model, but have a resolution of 512x512. You need to use 1 megapixel sizes like 1024x1024.

[onyx solar]

All I can tell you is that you have the wrong one. If the model is XL, you need the XL VAE, if it's 1.5, you need a 1.5 VAE. If you're set to automatic, try changing it to the matching VAE specifically.

[distant topaz]

It didnt do much

[onyx solar]

Your schedule type is set to automatic in the previous screenshot. You should try some other combinations of Sampler/Scheduler as well.

[onyx solar]

Also, not familiar with that checkpoint...where did you get it and are you sure that's not a LoRA?

[distant topaz]

LoRA? I followed a guide on youtube and got what they said. Got it from CivitAI

[onyx solar]

Yeah, but what link, specifically?

Nevermind, I found it.

That is a LoRA, not a full model.

You use LoRAs to influence other full checkpoints/models; you cannot use them as your checkpoint/model alone.

That means you likely have the file in the wrong folder.

That should go into the loras folder, not the checkpoints folder.

On top of that, this LoRA is meant for use with Pony Diffusion XL and that's an entirely different conversation about the differences with Pony versus other SDXL models and what it takes to prompt them correctly.

I'd suggest going back to YouTube in search of some other tutorials to get a more basic understanding and try again.

[violet swift]

any solution for this?

File "C:\Users\kevin\Desktop\ComfyUI\ComfyUI\custom_nodes\comfyui_controlnet_aux\node_wrappers\tile.py", line 18, in execute
from controlnet_aux.tile import TileDetector
ModuleNotFoundError: No module named 'controlnet_aux.tile'

[formal trout]

can someone help me install SUPIR?

[distant topaz]

I will do that. Dude stopped halfway through the guide, trying to push his patreon

[distant topaz]

Should it be those named embedding, checkpoint or XL?

[onyx solar]

You need to install the various model files for controlnet. Go lookup "controlnet model downloads".

[onyx solar]

Should what be named that?

[violet swift]

but i have the models for controlnet

[onyx solar]

You don't have the specific one it's looking for, apparently.

[violet swift]

but tile is a preprocessor not a model, I can't find that preprocessor folder

[pliant cipher]

Hi, I know that if I want to have my face appearing on SD, I should train a lora with some picture head shots. How can have the rest of my body trained as well? Haven't found this anywhere.

[onyx solar]

In your custom_nodes\comfyui_controlnet_aux\src\controlnet_aux folder, do you have the tile folder?

[onyx solar]

Seems like you should have what you need. I'd imagine the issue you added to the git will net you the answer.

[violet swift]

no one answers on github, damn it, this is my workflow

[onyx solar]

You've got to give it time for people to answer.

[violet swift]

F

[onyx solar]

🤷 💀

[violet swift]

If anyone here knows what's going on, I'd really appreciate it.

[dense tapir]

I get this nonetheless

[dense tapir]

This was with webui-user.bat

And I get this with webui.bat

[sinful needle]

nope. still not working

[ornate elk]

Can you show me the webui-user.bat

[dense tapir]

Hi!

[ornate elk]

The --ckpt-dir and the --update-all-extensions need to go in the same line behind --medvram-sdxl

Then open up a cmd and run
Pip cache purge

Then relaunch the webui-user.bat

[dense tapir]

In the SD folder?

[ornate elk]

Nope

[dense tapir]

Okay

no module 'xformers'. Processing without...

[ornate elk]

Can you show the full log again?

[ornate elk]

Try delete the venv folder and relaunch the webui-user.bat

[dense tapir]

@ornate elk Thank you, you're great as always

[ornate elk]

Np, does xformers works now?

[dense tapir]

Yes!

[ornate elk]

Perfect 🙂

[ornate elk]

Don't use tiled diffusion for upscale. Only enabled tiled vae.
Set the Resolution to a square resolution like 512x512 or 768x768
That will be the Tile size not the output res.
Then enable the Controlnet tile.
And at the bottom you activate the SD Upscale script.

Select an upscaler and generate

[dense tapir]

lol this guy might disagree https://civitai.com/articles/4560/upscaling-images-using-multidiffusion

[ornate elk]

Yea it can work too but mostly SD upscale script does the job

[dense tapir]

Yeah the scripts just seem superior to me

[livid fox]

Can someone help me? i just try to use "ControlNet" for poses.

[willow kettle]

Any tricks to get AMD working with SDXL? If I just drop in the SDXL base model it appears to generate but then at the end the process dies with

Exception Code: 0xC0000005█████████████████████████████████████████████████████████████| 15/15 [00:12<00:00,  1.17it/s]
0x00007FF9B7DE3064, C:\stable-diffusion-webui-directml\venv\lib\site-packages\torch\lib\c10.dll(0x00007FF9B7D80000) + 0x63064 byte(s), ?typeMetaDatas@TypeMeta@caffe2@@CAPEAUTypeMetaData@detail@2@XZ() + 0x6D34 byte(s)
0x00007FF9B7DE236C, C:\stable-diffusion-webui-directml\venv\lib\site-packages\torch\lib\c10.dll(0x00007FF9B7D80000) + 0x6236C byte(s), ?typeMetaDatas@TypeMeta@caffe2@@CAPEAUTypeMetaData@detail@2@XZ() + 0x603C byte(s)
0x00007FF9B7DE4807, C:\stable-diffusion-webui-directml\venv\lib\site-packages\torch\lib\c10.dll(0x00007FF9B7D80000) + 0x64807 byte(s), ?typeMetaDatas@TypeMeta@caffe2@@CAPEAUTypeMetaData@detail@2@XZ() + 0x84D7 byte(s)
0x00007FF9B7DE44AE, C:\stable-diffusion-webui-directml\venv\lib\site-packages\torch\lib\c10.dll(0x00007FF9B7D80000) + 0x644AE byte(s), ?typeMetaDatas@TypeMeta@caffe2@@CAPEAUTypeMetaData@detail@2@XZ() + 0x817E byte(s)
0x00007FF9B7DE3CE0, C:\stable-diffusion-webui-directml\venv\lib\site-packages\torch\lib\c10.dll(0x00007FF9B7D80000) + 0x63CE0 byte(s), ?typeMetaDatas@TypeMeta@caffe2@@CAPEAUTypeMetaData@detail@2@XZ() + 0x79B0 byte(s)

etc...

[ornate elk]

Sdxl should work with your card

What settings did you used?

In txt2img

[willow kettle]

Nothing special, DPM 3M SDE Exponential, 15 steps, 768x768

[ornate elk]

Okay

[ornate elk]

Can help you later, need to go now.
Make sure you have tiled vae extension and the sdxl FP16 fix vae

Also make sure you have 16gb RAM or better 32. when you have 16gb you need to increase your windows pagefile.
Also try downloading the latest zluda 3.8 files and replace the old ones

[solemn atlas]

what does that mean?

[willow kettle]

Thanks will try these in a few hours and report back

[sinful needle]

i set the batch size to 4 on animatediff and still getting the error

[unreal anchor]

I don't get it, SD tells me:
No SDP backend available, likely because you are running in pytorch versions < 2.0. In fact, you are using PyTorch 1.13.1+cu117. You might want to consider upgrading.

but if I do pip show torch in the cmd it tells me

Version: 2.3.0```

what am I doing wrong?

[lavish jacinth]

havent launched A111 in a while and i got this message when i loaded up today

venv "F:\Stable-Diffusion\stable-diffusion-webui\venv\Scripts\Python.exe"
fatal: detected dubious ownership in repository at 'F:/Stable-Diffusion/stable-diffusion-webui'
'F:/Stable-Diffusion/stable-diffusion-webui' is on a file system that does not record ownership
To add an exception for this directory, call:

    git config --global --add safe.directory F:/Stable-Diffusion/stable-diffusion-webui

fatal: detected dubious ownership in repository at 'F:/Stable-Diffusion/stable-diffusion-webui'
'F:/Stable-Diffusion/stable-diffusion-webui' is on a file system that does not record ownership
To add an exception for this directory, call:

    git config --global --add safe.directory F:/Stable-Diffusion/stable-diffusion-webui

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: <none>

[unreal anchor]

Probably that the resolution you choose was too big, or something in the process is asking too much for your gpu, I get that when I choose too much resolution or too much hires fix upscale, putting it lower makes it go away

[copper fulcrum]

Anyone know anything about this?

[solemn atlas]

do you think it has something to do with this

[ornate elk]

Out of vram

What's your txt2img settings?

[wary maple]

nothing in the drop down list even i just created an embedding file. Any idea?

I did press the refresh button btw

[solemn atlas]

it's this

[ornate elk]

Always set hires steps to 10 or it will error out

[willow kettle]

I updated ZLUDA, got the FP16 fix vae and the tiled vae extension. Do I need to change command line args from the guide?

[ornate elk]

No your fine with the guide ones

What you can try if you still get the error is to remove --medvram-sdxl

[willow kettle]

k trying a few tests...

[ornate elk]

For tiled vae extension, only enabled tiled vae. Not tiled diffusion

[willow kettle]

Ok I think I've got it working. Tiled VAE says the input size is tiny and unnecessary to tile. I can make it work with or without tiled vae (the images are slightly different with and without). I think my limiting factor was the pagefile, which I increased. (lol I am running with a paltry 8GB of system RAM)

[covert lotus]

@ornate elk I’ve followed all the steps and I’m doing my first generation is it meant to be like this-

Like blank with nothing happening

[solemn moon]

something updateed and now all my images are like heavy blurry in the background

Steps: 20, Sampler: DPM++ 2M, Schedule type: Karras, CFG scale: 7, Seed: 2093861087, Size: 1024x1024, Model hash: 31e35c80fc, VAE hash: 63aeecb90f, VAE: sdxl_vae.safetensors, ControlNet 0: "Module: none, Model: control_v10e_sdxl_opticalpattern [1e4e1607], Weight: 1.1, Resize Mode: Crop and Resize, Processor Res: 512, Threshold A: 0.5, Threshold B: 0.5, Guidance Start: 0.13, Guidance End: 0.91, Pixel Perfect: True, Control Mode: Balanced", Refiner: sd_xl_refiner_1.0 [7440042bbd], Refiner switch at: 0.8, Version: v1.9.3-1-gddb28b33

[willow kettle]

The very first time you generate with AMD using the guide, it can take up to 40 minutes. One-time cost.

[ornate elk]

Yes looks good

[covert lotus]

Ah ok, I knew it took awhile I just thought there might be some indication that it was working

It’s fine to use a different checkpoint rather than the base on tho right

I’m using animanga v31 for my first generation

[ornate elk]

Ah nice. If you upscale then always enable tiled vae

That helps reduce vram and generate faster

[ornate elk]

You can use any checkpoint (model) you find on Civitai.com
animagine v3.1 is sdxl based so it needs a base resolution of 1024x1024

[covert lotus]

Ah is that why it’s maybe not doing anything at all?

[ornate elk]

No it will load

[covert lotus]

Ah ok sweet ty

[ornate elk]

First gen can take very long

[covert lotus]

Also I do have another question regarding sd as a whole

WHERES ALL MY FILE STORAGE GOING😭😭

[ornate elk]

Later open up a cmd and run
Pip cache purge
You will get 5-10gb back

Temp files from the install

[covert lotus]

Ah great ty

[cosmic furnace]

All of a sudden. When trying to make a bigger image in img2img . This error pops up -
NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check

Turning on this Upcast to float32 seems done nothing

[willow kettle]

Well I seem to be up and running again. @ornate elk you are a paragon of excellence

[covert lotus]

I also notice when I’m using the program I lose like 50 gigs but when I close it they come back is that normal

[ornate elk]

On C?

It could be related that you used an sdxl model. It builds up a huge pagefile that is located on C drive

My pagefile is 40gb

[cosmic furnace]

and now it says out of Vid?memory - OutOfMemoryError: CUDA out of memory. Tried to allocate 3.38 GiB. GPU . o.0 And with medvram give that error -_-

when 1280-1024 o.0

The first model-check i have any problems with 🤔

[lavish jacinth]

so im trying to reinstall A1111 from scratch and i installed it but when i launched the bat file i got this

[covert lotus]

Yeah I think so, cus last night before I installed anything I had 890, now I have 806 and when it’s running I have 760

[lavish jacinth]

i got pip to update but not torch

[ornate elk]

Did you get the first image meanwhile?

[wind thunder]

TypeError: 'NoneType' object is not iterable

i get this when i try to generate an image on forge

some people says its because controlnet idk

Traceback (most recent call last):
File "E:\SD-FORGE\stable-diffusion-webui-forge\modules\call_queue.py", line 57, in f
res = list(func(*args, **kwargs))
TypeError: 'NoneType' object is not iterable

the cmd error

[ornate elk]

What settings?

[ornate elk]

Install python 3.10.11 64bit
Then delete the venv folder.
Then open up a cmd and run
Pip cache purge
Then launch the webui-user.bat

[cosmic furnace]

Yu mean this ? set COMMANDLINE_ARGS= --use-zluda --update-check --skip-ort --no-gradio-queue --ckpt-dir

if --medvram-sdxl added also says out of mem often. when txt2img and upscaling can handle more less 1280-960 . Weird but i guess badly optimised model dunno

[covert lotus]

Yeah it’s all working now

Just worried about my ssd getting clogged

@ornate elk I think it crashed my pc Lmao

[ornate elk]

xD with sdxl you have to be carefull when using hires fix

if you use hires fix, set the hires steps to 10

[covert lotus]

I don’t think I had hires fix enabled

[ornate elk]

make sure you use tiled vae when upscale sdxl

[covert lotus]

I just had Euler a 28 sampling steps with the Lora file and animangaxl

[ornate elk]

okay, strange, also make sure your not running any heavy programm (games or Wallpaper engine) in the background

[covert lotus]

Just quit wallpaper engine 😭

[ornate elk]

WP Engine is extremely buggy when using SD. it can freeze or crash your PC when both are running

[covert lotus]

I recently upgraded my pc and kept getting crashes until I put my old cpu back in and it didn’t crash since, maybe it was power consumption after all

[ornate elk]

strange, what was your CPU?

[covert lotus]

I had a 3600 (what I’m using now) I changed to a ryzen 7 5800x

It kept crashing until I put my old one back in

So it was either psu or cpu related

[ornate elk]

oh did you also updated your mainboard bios?
Thats something you should do when switching to a higher gen

[cosmic furnace]

oh i didnt) indeed ;o

[rugged latch]

does sum1 mind helping me ;-;

im trying to install TripoSR and..i dunno wut im doing ;-;

[covert lotus]

Yeah I did

[ornate elk]

Oh but the AMD CPUs don't need much power, so idk why it crashes

[vocal flume]

Hi hi
I need some help. I am trying to generate images localy. But they keep making just a dark grey image

[ornate elk]

Hey, what's your GPU and what's in your webui-user.bat?

[vocal flume]

GTX 1050ti

For my .bat
git pull
@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--xformers --precision full --no-half --medvram --opt-split-attention
set OPTIMIZED_TURBO=true
call webui.bat

[ornate elk]

remove --precision full --opt-split-attention

[vocal flume]

Thank you
Will try

[ornate elk]

then make sure the model is a model and not a lora. you can check this by looking at the file size. if its larger than 2gb its a model. if not then its a lora or something else

[vocal flume]

Understood. Will also check on the size. I think i am using a lora then by accident

[cosmic furnace]

Sd next seem to not have this issue o.0

with this model at least... So mb optimal to have both 🤔

jpg weights alot less. How do i make usual auto also end up in jpg not png? Why png default anyway hm?

[unreal anchor]

Nobody answered me so I'm gonna ask again:

I don't get it, SD tells me:
No SDP backend available, likely because you are running in pytorch versions < 2.0. In fact, you are using PyTorch 1.13.1+cu117. You might want to consider upgrading.

but if I do pip show torch in the cmd it tells me

Version: 2.3.0```

what am I doing wrong?

[rugged latch]

i feel chu

:3

i had said dat

for ur issue tho

u need to update Python

torch is fine

you need Python 3.10. 6

and ur version is 1.13

so, download Python 3.10.6, (whish is the last update stable can use), and install dat :3

then open webui-user.bat, and update the (python= "") line with the new update path :3

and done

[rugged latch]

hope dat helps >:3

[ornate elk]

Hey, what's your webui version?

[unreal anchor]

Mmmm how do I check?

[unreal anchor]

Let me check 🙂

[unreal anchor]

See, I just did python -v on the cmd and it says I have 3.11.2

Maybe SD doesn't know?

ah, my webui user bat has "set PYTHON="
and it doesn't have anything, what do I have to put, the path to the latest Python exe?

[ornate elk]

Python 3.11 is not supported

[vocal flume]

Still not working. This time it did show a preview, but when it hits 100% it goes back to grey

[viscid gate]

what resolution am i supposed to use for sdxl

[ornate elk]

whats the model you use?, can you show your txt2img settings

[unreal anchor]

OOF

[ornate elk]

1024x1024 is the trained resolution

so anywhere around that

[viscid gate]

thanks is that before hires fix or after

[ornate elk]

check the bottom, of your webui in browser for the version

[ornate elk]

before

[unreal anchor]

I do have python 3.10 too

[viscid gate]

ok any other settings i shoujld look out for cause my images look kind of strange

[unreal anchor]

ok, let's wait till it opens haha

[vocal flume]

I am currently trying
Anime Dreamer Mixed

And those are my settings. I changed the resolution just to try testing if it was just mad the resolutuion was too low

512x512 also did that

[ornate elk]

depends on the model, some models need lower steps and cfg, also try a resolution of 768x1024 for portrait or 1024x1280

[ornate elk]

thats resolution is to hig hfor an 1.5 model

[vocal flume]

I see
But also happened with 512x512
Actually i cancelled that with higher resolution

[viscid gate]

ok ive noticed loras really break my sdxl generations, how can i fix this

[ornate elk]

can you show an example?

also what does the cmd shows?

[ornate elk]

what breaks?

[vocal flume]

Nothing
That's the weird part. Shows 100% then does nothing. Then if i try to generate another art it tries to generate it

[fair oxide]

any tips on how to get a girl to sit indian style?? I use pony, so its quiet difficult to use controlnet, it crashes my pc when using controlnet openpose on pony models, not sure if there is a workaround to that

[unreal anchor]

so, I now have python 3.10.6

version: v1.9.3  •  python: 3.10.6  •  torch: 1.13.1+cu117  •  xformers: 0.0.23.post1  •  gradio: 3.41.2  •  checkpoint: 88967f03f2

it still won't work

Launching Web UI with arguments: --xformers --no-half-vae --api
WARNING:xformers:WARNING[XFORMERS]: xFormers can't load C++/CUDA extensions. xFormers was built for:
    PyTorch 2.1.2+cu121 with CUDA 1201 (you have 1.13.1+cu117)
    Python  3.10.11 (you have 3.10.6)
  Please reinstall xformers (seeblahhh blah blah````

the thing is, I tried upgrading pytorch and reinstalling xformers

but nothing happened

doesn't seem to work

[ornate elk]

do you have the dreambooth extension installed ?

that can cause such issues

[unreal anchor]

I don't even know what that is

nah I checked and I don't

[ornate elk]

okay then delete the venv folder and
then open up a cmd and type pip cache purge
then relaunch the webui-user.bat

[unreal anchor]

uhh... where is the venv folder : $

[ornate elk]

Inside the stable-diffusion-webui folder

[ornate elk]

Can you make a screenshot of your txt2img settings when getting the grey image?

[vocal flume]

Model loaded in 10.5s (load weights from disk: 0.9s, create model: 0.5s, apply weights to model: 8.5s, calculate empty prompt: 0.4s).
100%|██████████████████████████████████████████████████████████████████████████████████| 20/20 [00:28<00:00, 1.42s/it]
Total progress: 100%|██████████████████████████████████████████████████████████████████| 20/20 [00:27<00:00, 1.39s/it]
Total progress: 100%|██████████████████████████████████████████████████████████████████| 20/20 [00:27<00:00, 1.40s/it]
Is all that shows after the images goes grey

[ornate elk]

Can you show the whole cmd when launching the webui-user.bat?

[rugged latch]

oh i had that happen before

i was getting black images

[lavish anvil]

How to create an image?

[rugged latch]

i had found it was bc i had accdentaly enabled something like, "make pictures tilable"

after disabling, pictures came out appropriate

[vocal flume]

Thanks
But tiling is disabled already

[ornate elk]

Try change --no-half to --no-half-vae then also add --opt-sub-quad-attention

[vocal flume]

Same result

[lavish anvil]

Help? I don’t know how to use this app and would like to know where to ask for help?

[vocal flume]

Maybe i added wrong?
git pull
@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--xformers --no-half-vae --medvram --opt-sub-quad-attention
set OPTIMIZED_TURBO=true
call webui.bat

[ornate elk]

Looks correct

Can you try an other model?

[vocal flume]

Gonna try that. Downloaded another one

[ornate elk]

Which app?

[lavish anvil]

Isn’t this an app?

I want to create an image on this discord “app?”

[ornate elk]

Discord is an app and programm.
Stable diffusion has an Bot on this server called Artisan.
Checkout #artisan-faq on how to use it

[lavish anvil]

Thank you!

[ornate elk]

If you have a good PC with a GPU you can also create images localy

[vocal flume]

Thank you
It now works
Looks a bit scuffed, but could be the prompt or something

[unreal anchor]

oke let's see

[rugged latch]

Yay

might have been that it was currupt

if ur pics are a bit scuffed still, u can either try, lowering the res, widthxheight, or increase sample steps :3

or, try a diff sampler uvu

[vocal flume]

The file was probably corrupted. Tried another model and it made a great image

[covert lotus]

probably faulty cpu then

[regal hawk]

hey ! Someone knows why I can't train Loras some already finetuned models from civit ai when using kohya ? Or is there a way to fix it ?

[stark parrot]

can someone tell me how to manually update fooocus? it dose not update automatically. i dont believe its a ''fork'' . in the cmd window it says im using 2.1

[sinful needle]

Did. Still not working. This is so confusing. Before my factory reset all was working, now I'm back to square one despite just copying the repos entirely without changing it.

I tested it again at just 512x512 and it works... okay-ish. 768x768 gives me that big error.

[slow breach]

Hello! How can I get started with image generation here on discord ?

[rugged latch]

wha

[rugged latch]

wut do u mean

stable diff, isnt on discord i dunt think <v>

its a, python thing u launch on ur pc

[ornate elk]

Hey, checkout the #artisan-faq channel

Thats the bot to generate images here

[rugged latch]

ahh

[slow breach]

But how are images generated, for example in this channel: #1237459938901491852

[rugged latch]

neat

[onyx solar]

If you go through the link that you were given and read, you will eventually get to the point where you have the information you seek. But there is no free image generation on the server, just so you know.

[hollow ocean]

I decided to reinstall due to some errors and now I am having a UI issue and I am not sure if it something related to the install or something I messed up in the settings I am seeing this instead of the normal search and sort/refresh and it doesn't work to refresh the lora/checkpoints

[sinful needle]

Did batch of just 4 for AnimateDiff. Still not working. This is so confusing. Before my factory reset all was working, now I'm back to square one despite just copying the repos entirely without changing it.

I tested it again at just 512x512 and it works... okay-ish. 768x768 gives me that big error.

[dense tapir]

@ornate elk Hey! Sorry for pinging you about non-AI stuff, but would you happen to know somebody or maybe just an entire community dedicated to helping with After Effects tech support?

[wicked geode]

anyone have good wav2lip, my version is outdated and kinda blurry D:

[solemn moon]

🫠

[wicked geode]

Does R-ESRGAN 4x+ upscale on a 1:1 upscale usually slow?

doing a batch process with 101 images

[quasi ginkgo]

is it normal for A1111 to be using 22 gigs of ram

[grim iron]

I get this error while using dyncamic promts can anyone help? WARNING:dynamicprompts.samplers.utils:No values found for wildcard 'rain'

[oak wing]

Hi I am new to here can anybody tell how to use Stable Diffusion here ?

[grim iron]

here is a guide

[oak wing]

?

[bleak plover]

hey guys, both in Forge and auto1111 i am getting weird error when i try to create new embedding. Is there a known bug? Here's my error message

[upbeat gyro]

Hello, Whenever i try to run SD Forge UI Web UI, i get this error, How can i fix? I really need help since i want to make an animation

[ornate elk]

Make sure the webui is whitelisted in any browser adblocker

[ornate elk]

If your using sdxl models then yes

[ornate elk]

Hey, make sure your nvidia driver is updated

[upbeat gyro]

Ill try that, thanks for now 😄

Its already at the latest driver

[ornate elk]

Okay and your python version is 3.10.6 ?

[pearl forge]

Hello, I have body measurements and a selfie, and I would like to create a 2D avatar based on these two inputs. Is it possible to do this using Stable Diffusion, ComfyUI, or do you have any other suggestions?

[upbeat gyro]

Yes

[ornate elk]

Okay, was that with batch size 16 or 4-8 ,?
What motion module did you used?

[ornate elk]

Nope idk sry

[ornate elk]

Idk what you mean by body measurements but if you mean proportions, then yes you can do that in comfyui or Auto1111

[ornate elk]

What's your GPU?

[upbeat gyro]

Its an NVIDIA RTX 3060 12GB 2x OC

MSI

[pearl forge]

Body measurements like height, weight, neck, chest, etc - a lot of detail properties 😄

[ornate elk]

Ah okay, with numbers you can't do much in Stable-diffusion.
But you could create an 3D figure based on these settings. And with that 3d figure you can work in stable-diffusion using Controlnet (depth, normal, canny, openpose)

[pearl forge]

For instance: male with a height of 180 cm, weight of 80 kg. The body type is Triangle with an average build and relaxed posture. Please include the following body measurements: calf - 36.18 cm, knee - 37.41 cm, neck - 38.13 cm, ankle - 27.14 cm, bicep - 30.77 cm, chest - 97.78 cm, thigh - 59.25 cm, waist - 89.13 cm, wrist - 17.44 cm, abdomen - 94.42 cm, forearm - 27.80 cm, low hips - 100.57 cm, and high hips - 90.29 cm. For the linear measurements, please include: neck - 12.82 cm, waist - 29.42 cm, inseam - 76.59 cm, outseam - 110.41 cm, shoulders - 43.41 cm, jacket length - 80.72 cm, sleeve length - 62.82 cm, back neck height - 153.86 cm, upper arm length - 35.24 cm, inside leg height - 82.94 cm, and shoulder length - 15.68 cm. The male should have short brown hair and wear a fitted t-shirt and jeans. Please place him on a neutral background and position him with a slight smile and arms by his sides.

[ornate elk]

Yea Stable diffusion won't do that. It will ignore the numbers.

But as I said you may can create a 3d figure based on that with Blender or https://posemy.art/

[ornate elk]

Okay, did you installed Automatic1111 before forge?

Or only tried forge

[upbeat gyro]

Yeah, i just ran webui-user.bat and everything started to do the work, but github says to only run update.bat and run.bat

I have 2 Automatic1111's, One for dreambooth, and one for image generation

[ornate elk]

Ah okay

Thats good because dreambooth breaks often

[pearl forge]

I want to do it without manually, Is it possible?

[upbeat gyro]

Yeah i know, dreambooth doesnt allow image generation on my 1.8.0 repository, so i have 1.9.3 for that, and the 1.8.0 one just creates ai image checkpoints for me

[upbeat gyro]

But im now trying to run webui-user.bat in the webui folder to hope that works, i won't set my hopes high though

[ornate elk]

If you want a clean and correct install then checkout my install guide here:
#🤝｜tech-support message

It also covers the best launch args for performance

[upbeat gyro]

Thanks 😄

Traceback (most recent call last):
File "E:\Forge\webui\launch.py", line 51, in <module>
main()
File "E:\Forge\webui\launch.py", line 47, in main
start()
File "E:\Forge\webui\modules\launch_utils.py", line 541, in start
import webui
File "E:\Forge\webui\webui.py", line 19, in <module>
initialize.imports()
File "E:\Forge\webui\modules\initialize.py", line 35, in imports
import gradio # noqa: F401
File "E:\Forge\webui\venv\lib\site-packages\gradio_init_.py", line 3, in <module>
import gradio.components as components
File "E:\Forge\webui\venv\lib\site-packages\gradio\components_init_.py", line 1, in <module>
from gradio.components.annotated_image import AnnotatedImage
File "E:\Forge\webui\venv\lib\site-packages\gradio\components\annotated_image.py", line 13, in <module>
from gradio.components.base import IOComponent, Keywords
File "E:\Forge\webui\venv\lib\site-packages\gradio\components\base.py", line 20, in <module>
from fastapi import UploadFile
File "E:\Forge\webui\venv\lib\site-packages\fastapi_init.py", line 7, in <module>
from .applications import FastAPI as FastAPI
File "E:\Forge\webui\venv\lib\site-packages\fastapi\applications.py", line 15, in <module>
from fastapi import routing
File "E:\Forge\webui\venv\lib\site-packages\fastapi\routing.py", line 22, in <module>
from fastapi import params
File "E:\Forge\webui\venv\lib\site-packages\fastapi\params.py", line 4, in <module>
from pydantic.fields import FieldInfo, Undefined
ImportError: cannot import name 'Undefined' from 'pydantic.fields' (E:\Forge\webui\venv\lib\site-packages\pydantic\fields.py)

Now im getting this error, sorry for spam

[ornate elk]

the AI cant help you with body measuremtns, see this: first is your prompt, second image is without alle the numbers and cm stuff:

[ornate elk]

thats a known error that comes from controlnet

[upbeat gyro]

Oh, how do i fix it?

[ornate elk]

Normaly in Auto1111 it can be resolved by doing this:
Go into the stable-diffusion-webui folder. Click in the explorer bar (not search bar)
Type in cmd and hit enter.

Run the three commands one by one:

venv\Scripts\activate pip install albumentations==1.4.3 pip install pydantic==1.10.15

Then relaunch the webui-user.bat

idk if it will work on forge too

[upbeat gyro]

Ill try that, Thanks for helping me 😄

[ornate elk]

np

wouldnt recommend using forge as its really buggy

with your card your better with auto1111

[upbeat gyro]

The only reason i use forge is because z123 and svd is built-in, and comfy ui is difficult to use for me

[solemn girder]

Cannot figure out how to update torch

Ive tried everything online but it just refuses to download and Im pretty sure its why Im getting NaN errors

[ornate elk]

can you show me a full cmd output when you launch your webui?

[solemn girder]

Sure

[ornate elk]

oh wow, that webui version is over a year outdated

whats your GPU?

[solemn girder]

Idk

Gifted pc

[ornate elk]

okay, then open up the taskmanager, go on the performance tab, then on the left click on GPU and on the right corner it should be the model

[solemn girder]

I would need to go up and look at it

NVIDIA GeForce GTX 1080

[upbeat gyro]

I think ForgeUI is working, ill problaly upload an compiled archive on huggingface to save people from stress

Yeahhh boiii, it succeeded :D, i got forgeui working

[ornate elk]

okay to update your webui do the following:
edit the webui-user.bat
at the line COMMANDLINE_ARGS= you add: --xformers --medvram-sdxl --no-half-vae
then in the empty Row below you add
git pull
then save and close.
then delete the venv folder inside the stable-diffusion-webui folder.
And finally launch the webui-user.bat

[ornate elk]

how did you fixed it?

[solemn girder]

Will try in a bit

[upbeat gyro]

I tried what you said earlier, but instead, i put albumentations==1.4.3 and pydantic==1.10.15 in the requirements_versions.txt file

[quasi ginkgo]

@ornate elk do you get compensated for your work here or are you just a good samaritan?

[ornate elk]

I don't get paid, but i got discord nitro.
I just want to help if I can 🙂
Some people asked, so i made a Donation Link in my Bio, if anyone wants to spend a bit.

[quasi ginkgo]

hell yea, I'll buy you a beer

[ornate elk]

Thanks, really appreciate that!

[covert lotus]

Yeah I wanna sort u out something too after I get paid

[solemn girder]

Its stuck downloading

[ornate elk]

Press space

You paused it by clicking into the cmd

[solemn girder]

Ah

[ornate elk]

Pressing space will unpause it

[ornate elk]

It still says version 1.4.1

How does the webui-user.bat looks like?

[ornate elk]

The webui-user.bat is the second one from the bottom

[solemn girder]

I was supposed to start that one?

[ornate elk]

XD yes and to edit it before