🤝｜tech-support

[fossil dew]

oh HDD i guess

interesting 🤔

It's lagging my PC a bit, and taking forever this is what makes me think my 2080 cant run XL 😂

[ornate elk]

can you check where your windows pagefile is stored?

[fossil dew]

how do I do that?

[ornate elk]

https://www.tomshardware.com/news/how-to-manage-virtual-memory-pagefile-windows-10,36929.html

there, check if its enabled only for the C drive

and not enabled for D or any other drive

then its okay

[fossil dew]

yep there it goes crashing 😂 lol ok gimme a sec and I'll scope it out

🤔

im a bit confused from here because the direction on the website are different from what you're saying from here

This is what the webpage is saying to do:

but im not sure I understand\i dont wanna fk it up

Sorry @ornate elk for pinging- Would you be able step me though this bit real fast?

[ornate elk]

ah there is the issue

the pagefile is on D

that slows your whole PC down

[ornate elk]

so click on the D drive and select None.
then click on C drive and set it to "system managed"
then reboot the PC

[fossil dew]

no page file?

[ornate elk]

For D

[fossil dew]

ok cool

[ornate elk]

and system managed for C

then everything should be mauch faster

[fossil dew]

got it 😄

ill see if that helps XL run 🥲

[ornate elk]

but the PC reboot is required

[fossil dew]

yep! Ima go do that rn ty ill be back if I run into any issues

[ornate elk]

m2 ssds arent to expensive anymore, so maybe upgrade to at least 265gb for the C drive or more

because windows and programms mostly prefer to store their stuff on C and you cant do much about it

[fickle zealot]

I have 512gb m2 but i made two partition

[ornate elk]

ahh okay

yea then try to give C a bit more back if thats possbile

[fickle zealot]

I fucked it xd

[fickle zealot]

Now i hope i didnt break my pc

💀💀💀

[ornate elk]

whats in your webui-user.bat?

[fossil dew]

It's saying the same thing as before and now its saying terminate batch job?

[fossil dew]

uhhhhh

wdym?

[ornate elk]

hmm looks okay, but the timm error is strange

is your gpu driver updated?

[fossil dew]

uh ill ask my hubs to help me with that

[ornate elk]

also make sure you have enough free space on the C drive

at least 15-20gb

[fossil dew]

updating rn

hrh i have exsactly 21gbs lol

ill clear up some space on (both) my drives 💀

ty for all the help CS1 😄

[ornate elk]

no problem 🙂

let me know if it works

[fossil dew]

not a (issue) its still loading, but is --no-half-vae something Im going to need to fix?

and its giving me the same read out as earlier

  warnings.warn(f"Importing from {__name__} is deprecated, please import via timm.layers```

it opened the web UI and now its saying:

```Interrupted with signal 2 in <frame at 0x000002933D8EA9E0, file 'C:\\Users\\thetr\\Python\\lib\\threading.py', line 324, code wait>
Terminate batch job (Y/N)?```

[ornate elk]

how much RAM do you have?

[fossil dew]

uh

Where do I find that 😂

[fickle zealot]

U are my hero

[fossil dew]

wait let me ask my hubs

16?

[ornate elk]

check the task manager / performance tab

if you have 16gb you need to increase the windows pagefile to make sdxl models work

[fossil dew]

yeah its 16gb

[ornate elk]

set it to custom, and to 16000min and 24000max for C drive. Then reboot the PC

[fossil dew]

back to the link you sent me 🥲 ok brb

also my pc is like QUICK SHOT now

thank you for that

[ornate elk]

nice np ^^

[fossil dew]

ok rebooting brb

launching webui-user.bat

[ornate elk]

always launch webui-user.bat

not webui.bat

[fossil dew]

i typed it wrong

💀

i got you ~

doing the same stuff

[ornate elk]

press N

[fossil dew]

it closed 😐

[ornate elk]

hmm

any old extensions installed?

[fossil dew]

oh!

i relaunched and it ran

😄

still gave me that "future warning" one tho

[ornate elk]

ah okay

then also check in the webui for extenion updates

[fossil dew]

I'm selecting the XL model in the checkpoint box to see if it loads or crashes

rip

[fossil dew]

This I dont know tbh

[fossil dew]

I think I updated this yesterday tbh

in the CMD prompt doing uh something git...ping git or something like that idr exactly

[ornate elk]

okay

[ornate elk]

any error now?

[fossil dew]

no but it's not loading

pc isnt lagging and it hasnt crashed yet

just taking forever

[ornate elk]

okay thats normal as the models are still on a hdd, give it a bit time

[fossil dew]

OMG IT LOADED

testing it now 🎉

omg its working 😭

[ornate elk]

perfect 🙂

[fossil dew]

I'm so fkin proud, I did it all (with your help) myself 😄 wooo!

ok ok Im going to go mess around with stable now!

Thank you so much CS1!

I guess I might need help with making the quality look good now tho

its uh

[ornate elk]

no problem, have fun! 🙂 for anime stuff try out some illustrious models

[fossil dew]

The one Im running is a illustrius model

[ornate elk]

wrong vae

[fossil dew]

OH GOOD THATS A EASY FIX

any suggestions?

[ornate elk]

this one:
https://huggingface.co/madebyollin/sdxl-vae-fp16-fix/tree/main

[fossil dew]

ty ty

[fossil dew]

man hugging face is hard for me to understand 🥲 uh

[ornate elk]

there is a small download button

[fossil dew]

oh there it is!

into the VAE folder?

Do I have to select the VAE somewhere in stable diffusion or just reboot the python thingy

"web user ui.bat" or whatever

ohp! found it in the settings 😄

[ornate elk]

go into the settings, User Interface, Quicksettings
there add sd_vae
then hit apply and reboot

[fossil dew]

yep! thats where I found it!

Rebooting 😄

[ornate elk]

then you get an easy dropdown next to the model selection

[fossil dew]

o0o

😐 ima restart it and see if that fixes it cuz uh

[ornate elk]

yea reload ui button sometimes wont work cause of some extensions

then just restart the webui

[fossil dew]

face is a bit weird but I think maybe a different model might do some good- This is just the first one I downloaded 😂

😐

annnnnnnnnnnd it crashes

just when I think im done and its fixed, something else happens 🤡

🙃

its always something

[ornate elk]

Dont have to much tabs open and close background stuff like games or Wallpaper engine

[fossil dew]

I dont 😮

Stable and Discord are all I have running

[ornate elk]

Hmm

[fossil dew]

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`.

🤔

[ornate elk]

How much free space do you have on C right now?

[fossil dew]

i tried running it again and its giving me this plus the terminate batch job

uh ill look

im actually clearning my recycle bin rn so I'll reload, then check

[ornate elk]

Also open up a CMD and run
Pip cache purge

[fossil dew]

ok

30gb on C

184 on D

[ornate elk]

Okay should be good then

Pagefile uses disk space

[fossil dew]

should be- I'ma try reopening webui and see if it crashes again

🤷‍♂️

[ornate elk]

Yep

[fossil dew]

Well- It seems to be working for now if it crashes or does some funky stuff again I'll be back 💀

Do loras have to be XL loras to work with a XL model? is that even a thing?

[ornate elk]

Yes exactly

[fossil dew]

could explain why my images look a lil funky- awesome ty

[ornate elk]

Do you have an example?

[fossil dew]

uh 😐 just what I posed before with the faces being a little goofy

[fossil dew]

face/hands lookin' fucky

The lora I have on for her two tone hair, i was assuming because they're 1.5 that it could have messed with image gen quality

[ornate elk]

Yea dont add them, but normaly they just won't work

Also sdxl requires a higher resolution

Try 768x1024 for example

And then use adetailer for the face fix

[fossil dew]

I have adetailelr on- so we will try the different res and see what happens, I have also found XL versions of the LORAs I was running so I'ma try those too

and ofc I put the LORAS in the LORAs file and they're not showing up 🤡 I'll figure that out on my own- but jeeze it's just 1 thing after another

[ornate elk]

They only show up if a compatible model is selected

Sdxl for sdxl and so on

[fossil dew]

Yea hthey were XL but they weren't showing up

its ok tho because I still am getting split colored hair with the words (two tone hair, split color hair, pink hair, black hair) so I dont exsactly need the LoRa

Stable looks SO much better, the quality is AMAZING and my PC is running like butter

You're the best- a gold star for you ⭐

[ornate elk]

Thanks, and yep sdxl quality is awesome.
Also checkout illustrij, its perfect for semi realism:
https://civitai.com/models/1025051/illustrij

For anime also checkout Nova Anime Xl

[fossil dew]

https://civitai.com/models/827184/wai-nsfw-illustrious-sdxl

This is the model Im running

[violet harness]

I filled in a form on Stable Diffusion’s website and got the following message yesterday and today and still have not received any welcome email in my inbox or spam folder:

“Your Stability AI Community License has been successfully activated. Keep an eye on your inbox for our welcome email, filled with everything you need to get started and connect with the community.”

Is the form to email system down? Any advice?

[ornate elk]

Thats one of the best

[ornate elk]

Hey check #🗣｜artisan-support-feedback
Or #artisan-faq

This channel here is no officiall support for there paid plans

[maiden frost]

My current version of ComfyUI (runpod SDXL) is not showing rendered images in the tray in the bottom… That pane just isn’t visible for some reason. Any ideas?

[drifting geyser]

any good worfklow for wan that has quality? one of them got destroyed by updates and now I am left without workflow thanks!

[fossil dew]

is there any way to make XL make images faster? Or you think Im capped out- @ornate elk? Remember im running a RTX 2080 with 16gb ram (whenever you get the time to see this)

[misty oxide]

Hello. I have a very important question. If I want to create my own comics rot commercial thing by Stable Diffusion, may I have all right for those pictures after generation?

[crisp apex]

Hey all, I am suddenly having an issue with DS Forge, it won't launch and Im getting this error

TypeError: argument of type 'bool' is not iterable Exception in thread Thread-30 (webui_worker): Traceback (most recent call last): File "threading.py", line 1016, in _bootstrap_inner File "threading.py", line 953, in run File "C:\Users\*\AppData\Roaming\StabilityMatrix\Packages\Stable Diffusion WebUI Forge\webui.py", line 101, in webui_worker app, local_url, share_url = shared.demo.launch( File "C:\Users\*\AppData\Roaming\StabilityMatrix\Packages\Stable Diffusion WebUI Forge\venv\lib\site-packages\gradio\blocks.py", line 2446, in launch raise ValueError( ValueError: When localhost is not accessible, a shareable link must be created. Please set share=True or check your proxy settings to allow access to localhost.

Anyone know how I can fix this

It seemed to pop up after I download Visual studio

Well, I seemed to have fixed it by deleting some extensions... kind of odd, but it's working again

[ornate elk]

Which settings? Sampler, resolution and steps?

[fossil dew]

@ornate elk - I know I can make the width and Hight up to 1024 but I like the original setting uwu

that and I be genning stuff right- and all of a sudden:

its done this to me 3-4 times now

lets me gen a bit, then errors out

goddamn it- did it again >:[

only got to make 1 img before it fizzled out on me

did it again, made 1 image and the connection errored out

[ornate elk]

Set hires steps to 10

[fossil dew]

ill give that a try

connection errored out after genning 1 image again

[ornate elk]

okay then you need to install the tiled vae extension

it can help using hires fix and not erroring out

[fossil dew]

How do?

in exstensions I look up "Tiled vae"?

[ornate elk]

click on extensions, then on available, then on load from

then search for tiled vae

install and restart

then enable only tiled vae

[fossil dew]

ffs

okay I'll have to reboot to do the thing -_- frustrate

Frustrating how it can be working fine from like 7 am to about 11 am and now all of a sudden its like "Nope, not gonna do it!"

When you say enable only tiled vae, do you mean dont use adetailer, refiner and hires fix?

[ornate elk]

no i mean dont enabled tiled diffusion

only the tiled vae function with hires fix and adetailer

[fossil dew]

oh or DemoFusion or whatever

awesome ok thank

ill let you know if it does something funky again 🙃

You've been a trooper dealing with me ⭐

this popped up

and now

bangs head on desk

[ornate elk]

which settings?

it says out of vram

maybe try upscale by 1.5 instead of 2

[fossil dew]

ok

[ornate elk]

and hires steps to 10

denois 0.5

[fossil dew]

well- I've made it past 1 .... nope. 😐

spoke too soon i guess

head desk - head desk

I'm gonna give it a damn break for now. -_- I'm getting too frustrated

I dont understand how it can work, then all of a sudden decide that somethings broken 😐

[ornate elk]

if oyu get an out of vram error, you have to also restart the webui to clear the stuck vram

[maiden frost]

can anybody recommend a super lightweight comfyUI RunPod setup?
Most these pod templates I literally can't even get the ComfyUI to load or it takes forever. Not sure what the issue is; picking faster GPU doesn't seem to make it any better

[fossil dew]

how does one do that

[ornate elk]

just closing the cmd and relaunching the webui-user.bat

[fossil dew]

Oh I'm silly I didnt read it right my bad

[fossil dew]

ok well I'm back at it- I'm going to try and run it again and see if it crashes again

wonder if restarting my pc would do anything 🤔

  warnings.warn(f"Importing from {__name__} is deprecated, please import via timm.layers",```

``` 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`.```

These two things seem important to fix 🤔

Applying attention optimization: xformers... done.``` this also seems important to fix 🤔

I tried to run stable, but it didnt even gen 1 photo- just connection errored out again.

[vocal burrow]

catching the train on the run but you can probably ignore those. Those are deprecation warning for devs, telling them they'll have to update their code in the future.

[fossil dew]

oh huh ok

Im just trying to figure out why Im connecton erroring and unable to use stable after its been working all morning

[vocal burrow]

Also not really an error, I'm gonna assume you just entered some parameters (probably copy pasted stuff from civitai I assume) and therefore it's trying to use the "vaeFtMse840000EmaPruned_vaeFtMse840k" which you don't have installed.

[fossil dew]

huh OH does that mean I have to go to the VAE thing in the settings and choose a new different one?

wow yep

Could that be what's making it connection error everytime I use it?

[vocal burrow]

you don't """have to""" it fell back using "none" parameter instead of vaeFtMse840000EmaPruned_vaeFtMse840k. "none" basically mean "use whatever VAE comes with the SD model you currently have loaded"

[fossil dew]

HUH Oki

[vocal burrow]

no relation to connection error (which I don't see any trace of in the snippet of logs you're sharing)

[fossil dew]

yeah its not showing up in the logs

the uh

web page

just says "connection error" and then refuses to gen anything else

I'll see if it does it again and send ss of everything

[vocal burrow]

for this, you're using a 512x512 resolution output with a SDXL model (those are made for native 1024x1024 output resolution)

[ornate elk]

have you made the setting change to get the vae dropdown in txt2img ?

[fossil dew]

uh

doesn't sound like something Ive done???

ok

[vocal burrow]

In the webpage, settings
-> User Interface -> Quicksettings
-> "sd_vae"
apply and reload ui

[fossil dew]

oh nice there it is

errored out mid gen

[ornate elk]

was tiled vae checked?

[fossil dew]

FFS

restarting web ui and trying again cuz it wont let me do anything

[drifting geyser]

how to make char turn around like this? https://civitai.com/images/63116466

[fossil dew]

ok- everything is clicked.

High res steps at 10 - upscale at 1.5
Tiled vae on, adetailer on

[vocal burrow]

The resource for this is listed in the very link you just sent

[drifting geyser]

i know but it doesn't make the char turn around

[fossil dew]

So I did the thing, it gen one image, crashed, now my entire screen is black and I can't see anything except for the mouse button and the FPS GPU CPU information in the top right corner

🥴

Actually had to get on my phone to tell you

[ornate elk]

ohh

[fossil dew]

Yeeeahhhhhh

Going to turn it off and turn it back on again

and im back 😐

I tried to take a SS but my screen went back too quickly

[ornate elk]

seems like upscaling + illustrious and lora and a lot of tags is to much for your 8gb vram gpu

[fossil dew]

that's what I was saying 🥲

I was like "I dont think my PC can run XL"

WHICH IS WEIRD

CUZ IT WORKED FOR LIKE 4 HOURS

smh

[ornate elk]

it can run sdxl but not upscale it

[fossil dew]

ok ill try that and getting rid of the LoRa

🙃

that fixed it for now

we shall see

all that run around for the upscale and lora to be a problem 🥲 Im so sorry

it was the LoRa

[lucid yoke]

To be completely sure you are using the wan model right

Since i see you ask a lot of basic questions in general just checking

[fossil dew]

i can up the upscaler and its not crashing 🙃

Well atleast thats done with

sorry

and thank you again

[lucid yoke]

Since said resource works for me

[ornate elk]

no problem, glad it works now

[lucid yoke]

But yeah its a prompt thing you only need to tell it to do that rwynjix

[fossil dew]

me too 🥲 I dont gotta bug ya no more heh

yall have a good day- If I need anything else I know where to go 💀

[drifting geyser]

using i2v, i would like to try and make the character do a dance spin or a rotation that is smooth and natural, not quirky and disfigured or microwave rotation. Are we limited to do tis?

limitation on i2v at the moment?

[lucid yoke]

Hmm id say i2v is better for custom characters

[drifting geyser]

also I tried end frame, it basically disfigures my character even if it looks remotely the same, how do i keep consistent face?

would openpose be better ideea here?

[tepid crypt]

when it comes to train artstyle loras, should i add prompts to the images or not? what gives me results to have the closest artstyle in my generations?

[fossil dew]

I mean- It's lasting a whole lot longer before it crashes out so that's somthing 🙃

it says "connection error" on the web ui, but it stops half way though the loading bar and says "press any key to continue..." in the python-y thing.

Wish It would tell me what the error is 🙄

[glad fiber]

hey guys, what is the speed of stable diffusion on the default worflow from comfyui on the arc b580?

[true sphinx]

hey guys so sorry if this isnt SD related but i honestly cant figure this out and tried everything my brightness settings in windows 11 are missing. I cant like open settings and see a brightness bar or any other brightness settings. I have rolled back drivers, re installed drivers, re installed windows 11, gone into services made sure that display services was set to automatic multiple restarts of course. The main issue is that I cannot turn off adaptive brightness on my desktop. and my monitor keeps adjusting the brightness automatically.

[lucid yoke]

Its probably a monitor setting then. Maybe try googling the brand of the monitor or open its settings menu using its botton

[hidden furnace]

Hey guys, I just recently got a new pc (with an AMD gpu) and installed sd.next. Everything is working right, I can generate images without problem, but I can't use Control. I've read online and found out I had to download manually the models for it, however, even after downloading these files, the models doesn't show up in "controlnet", did I do something wrong?

[ornate elk]

hey, whats your gpu? and did you used the directml or zldua setup?

best is to not use SDnext as its overcomplicated.
Better use Auto1111 or Forge webui with Zluda

[fossil dew]

Is there like

a way to clear memory or make space for more or something

or is it just updatning the hardware in my PC

[fossil dew]

nvm lmfao I found like 81gb of sh* to delete in my downloads folder 💀 should do it

[ornate elk]

if you dont need it sure

[fossil dew]

yeah no, just old sims4 mods and old checkpoints/loras that I downloaded / never used anymore 😂

[surreal lily]

Is there any way to get comfyUI (or anything else for image gen) to work on an rx9070xt?

[ornate elk]

yes, check the first link of the pinned messages for the Zluda setup guides

[hidden furnace]

it's the 7800xt, I'm using zluda, and I've tried for days to use auto1111, but I keep getting the "runtime error torch is not able to use gpu" no matter all the fixes I've tried on the internet, SD.next was the only one I managed to get working

[ornate elk]

have you tried my install guide ?

its in the pinned messages the first link

[limber walrus]

Hi chads, hope you guys are all doing great

Quick question, how can i put those settings on for my generations ? That AIartist uses those settings but i cannot see them on my WebUI. Aswell as the " Override settings " if i don't send a png to the txt-to-image i don't know how to put it on.

Kind regards

[ornate elk]

hey, go into settings, user interface, quicksettings
there add:
sd_vae, clip_stop_at_last_layers, eta-noise-seed-delta, and rng

then hit apply and reload ui

[limber walrus]

Great ! Thank you CS1o 😄

[limber walrus]

Worked ! but can't find the " rng " option in the quicksettings, is it called a specific way? neither rn or rng is found hehe 😅

[ornate elk]

random number generator maybe?

[limber walrus]

hmm the only one it finds is " randn_source " but even " the word " random " is not found lol

[ornate elk]

hmm i mean its not a setting you have to change

its only if you want to produce exactly the same image as someone on a nvidia gpu, if you also have an nvidia gpu

[hidden furnace]

I did, I followed the guide exactly as written, still got the same error

[ornate elk]

then why not asked me or created an issue on github ^^

we can fix that error easily

[hidden furnace]

well then, if you can help me, I would love getting some guidance

[ornate elk]

sure can you show me the full cmd log of auto1111 ?

[hidden furnace]

ok, just give me some time, I can't do it right now

[ornate elk]

alright, just @ me when you can

[limber walrus]

Oh okay, thank you a lot. Also i see that that aiartist uses Stable diffusion xl, is there any documentation on how to start using it?

[ornate elk]

its a type of model (checkpoint), you can download sdxl based models from civitai.com

[limber walrus]

Oh i see, i thought it was different. I'm using SD 1.5

Said nothing xD i'm tired

it's just a different model

thanks again

[ornate elk]

no problem!

[hidden furnace]

@ornate elk I just discovered I'm extremelly stupid and can't read, lmao, I got the auto1111 working

[ornate elk]

Oh okay xD good that it worked now ^^

Nope out of memory means exactly that

Not enough GPU Vram to perform the task

What takes up vram:

• model size
• image resolution
• length of the prompt and negative prompt
• using loras
• using Hires fix
• using Controlnet

[fossil dew]

Thank you for the breakdown

So basically I can use XL but barely

[next nebula]

so running auto1111, getting closer but still so far away :/ I have downloaded these super nice loras, they are trained on flux 1.d . i grabbed a flux1-dev-bnb checkpoint to use with them, but not getting good results at all. what flux checkpoint should I use for these loras?

[fossil dew]

How can you tell what the model size is? Like- how big the download is in GB? Orrrr

[flat egret]

why even with the smaller strength, it keeps coming out like this?

[fossil dew]

I've always had an issue with feet unless theyre in shoes or socks 😐

[lucid yoke]

Image to image?

[flat egret]

inpaint

[lucid yoke]

I normally don't have a problem with feet if they are the image focus

[flat egret]

can i pass you the whole picture if you can test how it comes out for you?

[lucid yoke]

Hmm its a very small area with not a lot of pixels

[flat egret]

oh

[lucid yoke]

Oh I'm in bed still

[flat egret]

oh ok

[lucid yoke]

You could try

Upscale> then inpaint

[flat egret]

uhh

[lucid yoke]

More pixels to work with

[flat egret]

true, gonna try that, thanks

[lucid yoke]

And to be safe, if you inpaint, clean the prompt so it only has stuff that you want inpainted

[flat egret]

uhh, what do you mean to clean the prompt exactly?

[lucid yoke]

Remove everything that you dont want in the inpainted area?

Like i dont want a face there

[fair oxide]

guys, why during generation my nvidia gpu usage is still 0%
even though my gpu has been recognised by stable diffusion

[ornate elk]

Can you show a full cmd log?

[fair oxide]

wait before that, i am using the windows nvidia sd.webui from releases, not through git clone

[ornate elk]

Okay, have you also launched the updated.bat in there before the start.bat ?

[fair oxide]

i had during the first time, i dont think i need to use that every time right?

[ornate elk]

Yep only one time

[fair oxide]

bunch of absurd errors

but surprisingly opens

[ornate elk]

Why does it use conda lol, I thought it ships with python

Also it shouldnt be placed in Documents folder as its a system created folder, which can lead to permission issues

[fair oxide]

so where should i place it?

and should i not activate conda?

[ornate elk]

Yea dont use conda

If you dont need conda better uninstall it

[fair oxide]

does it create its own env or it will just pollute my main python

[ornate elk]

It creates its own venv

Best practice would be to not use the zip version and instead following my Install guide from the pinned messages

[fair oxide]

do i need to switch to python 3.10 before that or i can keep 3.13 as my main?

[ornate elk]

Nope you need 3.10.11 64bit

3.13 is not compatible with any ai stuff

[fair oxide]

that why i had used conda to switch to 3.10

but lets try switching without conda this time.

[ornate elk]

You can define the path to Python 3.10 by editing the webui-user.bat too

[stark vessel]

Guys i'm tring to do movements animations with this simple idle sprite, but i try to use the control net to move right, the image that comes out it's basic the same image i already have with 0 moviments change, and finally i try make it move in the direction it already facing, but nothing happen, anyone know what i'm missing?
I'm using SDXL, pixel-art-xl SoRA and thibaud_xl_openpose idk nothing about controlNet or anything advanced about AI genenrated images, i start today and could build some images but i want to make it usable to my spritesheets, so if i can't make it do all movements in one prompt i'm trying to do movement by movement, frame by frame

any guides? any help? anything would help a lot

[fair oxide]

is upgrading pip of python3.10 advisable or should i leave it as it is?

[ornate elk]

You can upgrade it

[fair oxide]

python3.10 done and now this is the output

i can see my cpu is in usage, probably nvidia is not being used

[ornate elk]

What's your gpu?
Also the webui is still in the Documents folder

[fair oxide]

4050
you didnt say where should i place my webui

[ornate elk]

Create a folder for it on C like C:\Ai

If you want to get it working correctly just Follow the install guide and dont use the zip variant

The guide also lists a few launch args for better performance

[fair oxide]

hmm i am watching

btw what will happen if i dont use medvram? I dont use any other thing while AI-ing

[ornate elk]

It then will generate images slower because it then can not adjust the vram usage to process larger models

Its needed for GPUs with 6gb vram or less.
It speeds up the generation of images with SDXL based models because those are 6gb in size.

[teal creek]

I now have a cartoon character image, and I want to edit this character, such as changing the hair color, adding items to the hands, changing the clothing color, etc. Which API would be more suitable?

[little bough]

Does anyone know the root cause of this? never had it before but noticed it today when genning, I hadn't updated anything or used any different models, any idea?

[next nexus]

Hello I cant find any stable diffusion API library for python. Any suggestions?

[hot briar]

@ornate elk when I put --use-directml it uses gpu but not upscaling, when I don't it doesn't use gpu but upscaling

[ornate elk]

What's your GPU?

Dont use directml if you can

[hot briar]

rx 7800 xt

[hot briar]

I wasn't but it only uses cpu if i dont use it

[ornate elk]

Oh

Can you show me your webui-user.bat and a full cmd log?

[hot briar]

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--use-zluda --update-check --skip-ort --no-half --skip-torch-cuda-test

call webui.bat

[ornate elk]

Remove --no-half --skip-torch-cuda-test

Then relaunch

And show me the cmd log

[hot briar]

okay launching

RuntimeError: Input type (float) and bias type (struct c10::Half) should be the same
get this when tried generate

just now

log after got error

[ornate elk]

i see controlnet stuff in there? what do you try to do exactly?

please try to gen a normal 3ß steps, euler a, 512x512 image

and which model do you use?

[hot briar]

@ornate elk

these parameters

[ornate elk]

can you try without hires fix, lora and make sure controlnet is not selected

[hot briar]

same

[hot briar]

log after try generate without lora, hiresfix and controlnet

[ornate elk]

okay, delete the venv folder and relaunch

[ornate elk]

did it worked?

[hot briar]

can't launch

[ornate elk]

relaunch

[hot briar]

okay

[ornate elk]

the numpy error is normal

[hot briar]

Compilation is in progress. Please wait...
waiting for now

[ornate elk]

sounds good

let it compile it can take bit

[hot briar]

thanks a lot it worked very well

[ornate elk]

no problem 🙂

[hot briar]

@ornate elk suddenly that happened, it wasn't even generating anything

[ornate elk]

if you used hires fix, always set the hires steps to 10 or you will run out of vram

[hot briar]

oh okay

[ornate elk]

if you want to upscale at larger res (with for example sdxl models) you also should enable the tiled vae extension. it can help a lot

[hot briar]

okay installing it

btw RuntimeError: Not enough memory, use lower resolution (max approx. 448x448). Need: 0.0GB free, Have:0.0GB free
i got that error

[ornate elk]

Can you share your txt2img settings when you get that?

[hot briar]

likethat

[ornate elk]

Makes no sense.
Are you running anything GPU heavy in the background like multiple Browser tabs, games or Wallpaper engine?

[hot briar]

no, only %3 of it being used rn and still have same issue

lol

[ornate elk]

Scam bot

[hot briar]

it was looking dangerous

[ornate elk]

Can you show me the full cmd log again?

[ornate elk]

try delete the .zluda folder. its inside the stable-diffusion-webui-amdgpu folder

then relaunch

[hot briar]

okay will inform

[fossil dew]

Since I can barely run XL, is there anything less powerful than XL but better than 1.5 that I might be able to try? Idk anything about flux or pony or any of these other things I hear about.

[ornate elk]

nope
you can only try Forge webui as its better optimised for gpus with 8gb vram or less

pony and illus are based on sdxl

flux wont run good on anything with less than 12gb

[fossil dew]

makes sense; gosh 1.5 is 💩 compared to XL it'd suck to go back.

Are there any directions on how to switch to forge?

[ornate elk]

you can check the pinned guide for an easy install

its just downloading an zip basically

you can also link the models so you dont need to copy them over

the guide also tells you how to do that

[fair oxide]

why cpu?

is my intel i9 13th so powerful that a GPU is nothing in front of him?

[hot briar]

it looks fine thanks

[hot briar]

uh

[ornate elk]

not enough memory while you didnt generated an image?

or have you deleted the prompt?

[hot briar]

used that thing

i dont know what it called

[ornate elk]

styles

but you said it worked before?

what changed?

[hot briar]

same when like that

[ornate elk]

try maybe restart the PC

[hot briar]

it was doing when hires off

but know doesn't do both situation

[hot briar]

okay restarting

[surreal lily]

So I was just following the guide for installing comfyUI with ZLUDA. I have installed rocm 6.1.2, downloaded the ZLUDA thing and unpacked it and added the things to the PATH etc. I skipped Step 3 as I have a 9070 XT, but when I launch the .bat, nothing happens

[fossil dew]

I see auto1111 and comfy but not forge?

[ornate elk]

can you try to gen a normal image?

after the restart?

[ornate elk]

you should install rocm 6.2 not 6.1

[surreal lily]

I did, still not working

[ornate elk]

can you show me your path settings?

[ornate elk]

but its there

[ornate elk]

looks correct

[ornate elk]

have you tried setting up auto1111 before or forge?

[surreal lily]

I actually did have auto1111 installed and running like 1 year ago using my old 3070

[ornate elk]

ah okay

but i mean with the new gpu with the guide

because if that works first we can easier fix comfyui

[surreal lily]

Nah, haven't tried it yet

Okay, I'll try that then

Should I try installing wiht DirectML or ZLUDA?

[ornate elk]

always zluda

[surreal lily]

Just tried launching A1111, the ui opened, but now it has downloaded some 1.7gb file and has been compiling something for the last 7-10 minutes

Is this expected?

nvm I got this error:

[ornate elk]

yes compiling is expected

[surreal lily]

?

[ornate elk]

relaunch and then try to egenrate a normal image

[surreal lily]

Does this take longer the first time I do it? Because it loaded the model, but

nvm it's doing something

[ornate elk]

yes the compiling will take a while

[surreal lily]

like 10min+?

[ornate elk]

yes

15-40 mins

more like 20

[surreal lily]

Ah, it works now

[ornate elk]

perfect

[surreal lily]

Also, quick question, Are checkpoints like image gen models (Like for LLMs where you have Gemma, Gemini, 4o, etc)or is it something else?

[ornate elk]

checkpoints are image gen models but similar to llm models

[glossy prism]

hey does stable diffusion offer a free model? i wanna try out in my app am making, i was using 3.5 large but now its asking for payment, so any free model so i can keep working on my app?

[languid sedge]

Yes, Stable Diffusion offers several free, open-source models that you can integrate into your application.

[lucid yoke]

Hes probably referring to the API

[glossy prism]

oh can u name some?

[glossy prism]

yes i am looking for api rn, when am done with making the app i do have plans to host some open source stable diffsuion model on collab or kaggle

[lucid yoke]

it depends on the hardware you have

if you can run it

[glossy prism]

online, on google collab or firebase

when i was asking chatgpt it told me, i can do that

[hot briar]

it does but not big resolution ones

[fallow veldt]

Hi, newbie here can I ask how to install SD? Any tutorial can you recommend? Truly appreciate. Thank you

[surreal lily]

So I just downloaded a checkpoint from civitai and now after I tried to load it, it started downloading "open_clip_pytorch_model.bin" is this normal?

[surreal lily]

Tried generating an image, got this error
return torch.embedding(weight, input, padding_idx, scale_grad_by_freq, sparse)
RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cpu and cuda:0! (when checking argument for argument index in method wrapper_CUDA__index_select)

---

[ornate elk]

Does a 1024x1024 work?
What was your GPU again?

[ornate elk]

Yep

[ornate elk]

How large is the the checkpoint file?

[surreal lily]

like 6.7GB

[ornate elk]

Should work then

Can you try relaunch and try again and then send the full cmd log?

[surreal lily]

Ah, I relaunched it and now it works apparently

[hot briar]

works, rx 7800 xt

@ornate elk worked with this settings for now

[ornate elk]

If you want to upscale sdxl models then you need the Tiled VAE extension first.

You then can test if it works then with hires fix and 10 hires steps

[hot briar]

I installed it but don't know how to use it

[ornate elk]

Only check the box where it says Tiled VAE

Leave it on default

That can prevent out of vram errors when upscaling at to large resolutions

[hot briar]

oh okay

didn't get any error today

[hot briar]

does it work with others?

[ornate elk]

Yes with every model

[acoustic hamlet]

hey fu

guys

my forge flux generates just black screen

[ornate elk]

Hey, do you have the VAE and the two clip models too?

[timber oak]

heya

does this mean my ai uses cpu?

if yes, how can i change that? i launched gpu.bat on my comfy

cuda cores are working

but it takes 32GB of ram

it is very slow

is it possible that my gpu is too slow for the model?

(i am using wan model to ai gen a txt2vid)

i am very new to video generation

and i was wondering how to optimise the process

there are websites that do it in 10 min

these are the settings the template came with

[vocal burrow]

it s using your gpu

[timber oak]

ok nice

[vocal burrow]

offload devices are what are used whenever there is not enough resources on your main device

not enough vram ? let s spill out to ram instead of vram and use cpu.

[timber oak]

ooh okay

so i have almost 12gb of vram + 32 gb of ram

thats being used by the ai

jeez

44GB of ram in total is quite a lot

flux can generate pic in 10 sec

[vocal burrow]

12gb of vram being actively used.
32gb of ram storing models for faster transfer to the gpu if they ever need to be (re)loaded.

[timber oak]

it uses a lot of ram

but its very quick

[vocal burrow]

cause loading from ram is faster than from your ssd or whatever

[timber oak]

ah i see

well yes

the wan model is absolutely massive

so what does that mean?

is everything normal?

do all videos take an hour to generate with wan?

i am still new to video stuff so i was curious if you have any tips for efficiency

[vocal burrow]

An hour is probably a bit much. there must be something wrong with your workflow. But I m not a wan guru either. I don t have a big enough gpu to effectively generate videos.

You can probably tweak it for better perfs.

[timber oak]

i am using the standart template

from comfy

[lucid yoke]

iirc using wan with 12 is "pushing it"

getting drastically different times though on my 5080 (9-15min depending on complexity)

then again 16gb vram

[lucid yoke]

Are you using the 720P model or the 480P?

[vocal burrow]

I d say browse around the web and see what other are doing with their workflows.
eg : https://comfyworkflows.com/workflows/5df9ee95-3bb7-4bbe-b385-fb0c24da324c

[timber oak]

480

[lucid yoke]

hmm im not a comfy guru but i do know those times are a bit unreasonable

[lucid yoke]

swarm handles it for me. im at work rn so i cant peep my workflow rn but ill @ when im home

[timber oak]

alright thanks 🫂🙏

[vocal burrow]

there s also probably some fp8 version of this one.

[timber oak]

whats the difference between fp8 and 16?

[lucid yoke]

oh good catch, yeah that could help or a gguf version

precision

is all i remember

[vocal burrow]

the format used to store numbers/data.

floating precision 8/16 bits

[timber oak]

my model is 16 but clip is 8

[vocal burrow]

the bigger the more precise, but the more vram it uses.

[lucid yoke]

FP8 uses 8 bits to represent numbers with reduced precision, allowing faster computation while saving memory bandwidth.

[timber oak]

would it help if i put the clip to 16 too?

[lucid yoke]

it would help to lower wan to 8

[timber oak]

okay

[vocal burrow]

and the quality does not scale linearly with the precision.

[timber oak]

you said you use swarm

is it better than comfy?

[vocal burrow]

so you re not missing necessarily that much by swapping it for an fp8 model.

[lucid yoke]

its basically the same so yes

comfy back end

warm front end

[timber oak]

atm i have linked comfy to my stability matrix

[lucid yoke]

swarm*

oh

[timber oak]

its easy

[lucid yoke]

yeah stability matrix wont work with swarm

[timber oak]

oh damn

why

oooh wait

[lucid yoke]

matrix keeps resetting the settings etc

[timber oak]

i dont use comfy inside stability

i have the extra model paths.yaml linked to stability folder

it makes everything 100000000x easier to manage

i use this for ai

its an m.2 ssd

which i linked to my pc using my pcie slots

i was scared that it becomes slower

but it actually runs very fast

[timber oak]

my only issue, is that when i try to organize my models into folders, sm uses a lot of ram

[next nebula]

can i install forge ai on the same pc that is running automatic1111?

[hot briar]

does kohya ss work on amd?

[vocal burrow]

yes no problem.

[next nebula]

TY

[ornate elk]

Nope but OneTrainer works

I made a install Guide for it too

[hot briar]

with zluda one right?

py scripts/install_zluda.py didn't work it says it can't find the path @ornate elk

[ornate elk]

yes

[ornate elk]

Can you show a screenshot?

[ornate elk]

You have to follow my Guide

https://github.com/CS1o/Stable-Diffusion-Info/wiki/Lora-Trainer-Setup-Guides

[hot briar]

git clone https://github.com/lshqqytiger/OneTrainer.git && cd OneTrainer && py scripts/install_zluda.py
I pasted what have I found in your guide

[ornate elk]

You have to use it as one command

[ornate elk]

Oh I see the issue

You have Python 3.13 installed

That won't work with any ai stuff

Uninstall 3.13 and then git clone it again

[hot briar]

tyty looks fine rn

[sick elk]

Hi sorry for the inconvenience, but i want to know how make content like red cat on tik tok animation ? (locally)

[hot briar]

Can I stop onetrainer, turn off pc, and continue to train after I turn on it?

[tulip venture]

I have a question. How long does it take to get a membership into Stability AI API?

I signed up last night and just wondering how long I may have to wait to hear back

I signed up for the Community Membership

[azure magnet]

Hi, should I install comfyui in a venv or with condra?

[ornate elk]

comfyui comes now in a all in one zip file no need for conda or venv.
But if your on AMD then make a venv

[limber walrus]

Hello everyone, hope you guys are doing fine. Little question here i want to do a test and train a model with 31 steps x 7 epochs, but i see that by default this setting " max train steps " is set as 1600. Do i have to put it to 0 ? i don't understand it's utility. Thank you in advance !

[fallen otter]

I'm trying to use Pose editor. I followed the instructions from a YT vid, yet the Pose Editor tab doesn't appear in my Diffusion Webui. Restarted as per instructed....nothing.

I installed ControlNet too (I think). I can use a pic to set a Pose, just not the editor which is what I really want.

Any help would b appreciated. I use Stability Matrix.

Seems quite a few ppl have had this issue.

[wicked nest]

I'm having an issue with installation. I Googled it and searched for solutions, but I'm still not sure what's happening. I’ve used Stable Diffusion before and had no issues installing it, but that was a long time ago. These are two different installations, both from the guide— the automatic one being at the top and the other one step-by-step at the bottom. Eventually, the top one becomes like the bottom one, but it keeps spamming an error saying "the time is incorrect," for whatever reason. However, the time is actually correct.

time isn't the problem tho eventually sorts it self out believe it's a server issue on that one.

But I've set the arguments correctly and it's still not working

[fathom sage]

hey guys, is rtx4070 enough to run wan 2.1?
I am having problems with system allocation

[fossil dew]

Im getting a new error out of nowhere >_>

It's been working just fine 🙄 Crashing still, but not as fast as it was.

Well- I closed it and opened it again and it seems to be working; but I'd still like to know what the heck that was all about.

[jaunty fossil]

hi does anyone know how to set x/y/z plot to let x axis have different lora and y axis with different weighting scale like this? appreciate any help!

[fossil dew]

Geralt!

[lucid yoke]

Swarm UI has a grid generator built in
https://github.com/mcmonkeyprojects/SwarmUI/blob/master/docs/Basic Usage.md#testing-like-a-pro-the-grid-generator

Though personally i haven't gridded before

[jaunty fossil]

thx i solved with using variables

[lucid yoke]

nice

[fallow veldt]

I have trouble use stable can someone help? Tqvm

[next pilot]

send your enquiry 👆above and get instant help

[fallow veldt]

@next pilot thank you so much

[next pilot]

Success is not the key to motivation. Motivation is the key to success. Keep pushing forward and never lose sight of your goals."

"Remember that every small step counts. Even the tiniest progress is still progress. Keep going!"

"Challenges are just opportunities in disguise. Embrace them, learn from them, and use them to grow."

"Teamwork makes the dream work. Together, we can achieve greatness and overcome any obstacle."

"Rise above the negativity and believe in the power of positivity. A positive mindset can drive you towards success."

"Failure is not the end, it's just a stepping stone towards success. Learn from your mistakes and keep pushing forward."

"Believe in yourself and your abilities. You are capable of more than you can imagine. Trust in your potential."

"Stay focused, stay determined, and never give up. Your hard work will pay off, and your dreams will become a reality."

"Take pride in the work you do. Your effort and dedication make a difference, and your contributions matter."

"Always strive for improvement. Be better than you were yesterday, and let your progress inspire others."

[little bough]

`has anyone had this message with reforge before, should i be concerned? my version is up to date

[ornate elk]

Hey can you show your cmd log?

And dont click on suspicious linkes from the Bot that answered you.
Its a scam

[fair oxide]

Do you need help?

[ornate elk]

Yes

[fair oxide]

I DM you

[idle bolt]

Is there a way in inpaint to mask a Person in Pictures with one click? Like Magic Mask in da Vinci resolve?

[ornate elk]

Try update the extensions

[wicked nest]

I've got everything working now but ever time it goes Idle, Like left it over night generating images endlessly it works (generated 1,000+, 1440 by 1600 images. however I stopped it and it times out then when I relaunch the web ui it starts to uninstalls everything that is cached and then reinstalls everything again how ever it says its uninstall older versions but those versions are not installs to begin with. - after this process it does go through Stable diffusion does work. But it's not making sense to me

• the auto update argument is not on either.

[ornate elk]

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

You can also provide the cmd log to check

[delicate urchin]

Trying a bunch of wan workflows and having issues, I keep getting these blurry results with WAN. Anyone knows what is up with it?

[wicked nest]

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=

call webui.bat

no arguments set I previously had --skip-torch-cuda-test but removed it as i got it working

I didn't take a image of it -

Nvida 3090 -

[ornate elk]

Ah okay then you need to add
--xformers

That makes the generation run faster and use less vram.

Also when you relaunch it shouldnt install everything again

If that happens pls show a cmd log

Also make sure the webui is not placed in a OneDrive synced folder like Desktop, Documents, Downloads.
As this cause massive issues.

[wicked nest]

its sitting in - C:/AI/{directory}

I received this after restarting

[ornate elk]

Ah your on an old torch version then.

[wicked nest]

oh

[ornate elk]

Open up a cmd and type
Pip cache purge
Then delete the venv folder.
Then relaunch

It will then install the latest compatible torch and xformers

[wicked nest]

Thank you it's working Didn't realize it was a old version? I used the latest version that's pinned. Maybe I somehow ended up going to some archive.. Had to be me doing something different, launched with zero issues.

[ornate elk]

If you installed the webui recently its strange that it was old.
But what also messes up torch is for example is the Dreambooth extension.
It breaks the webui, so never install it.

But good that it works now 🙂

[median trellis]

support

[foggy sail]

why everytime i open comfyUI it tells me to install python packages?

[south totem]

Is there a go to guide on how to instal stable diffusion? Last time I followed a random youtube video, it worked, but feels like was buggy.

edit: found the pinned post 😄

[fair oxide]

my stable diff, gives me no xformer warning.

should i manually install?

[fair oxide]

and one more thing, why these AIs dont work with optimus laptops, i had to enable MUX to make it use RTX

[south totem]

ok I went with Forge , after installing it told me that PyTorch is not setup for my RTX 5080, what and how do I install this properly?

it just links me to the locally page from pytorch

to be specific, the card supports CUDA sm_120, but the installed only supports up to sm_90

[ornate elk]

What's your gpu?

[south totem]

so, I understood so far that I need to update PyTorch, but how to do it properly? I don't wanna mess up my install by just doing things till it works.

troubleshooting tips says to delete the venv folder, I don't have that one

[lucid yoke]

ohh your on a 5080 same

well

theres gonna be some funky business your gonna hit

[south totem]

yay 😂

any guide on what to do?

[lucid yoke]

hmmm

well the thing i did is no longer in the swarm discord

but

i could find this (swarm/comfy use only)

insert salespitch for swarm 🗣️ video gen, frequent updates, active (good guy) dev etc

[south totem]

im not using swarm but forge though 😄
I have a rough idea what needs to be done, but not so familiar in the field to know how to do it properly.

I want to prevent blindly installing things that may confuse the system.
I have a freshly installed PC, so its the first install, so everything is very clean

[lucid yoke]

hmmm

well what i did was delete a few lines in a few files (trying to remember hard)

but swarm & forge is different. idk if it works on forge even yet

no promises but:
https://github.com/lllyasviel/stable-diffusion-webui-forge/issues/2601#issuecomment-2764238691

@south totem

some people are t alking about a fix here

though if you get stuck on this because they seem to be argueing in there

i got a working swarm instance (and the backend is portable) i can help with that one

till theres a official release/fix

[south totem]

ha! this seemed to have worked, after running both and starting the run.bat again, there is no error on Cuda

next question. I have the hint now "your device supports --cuda-mallox for potential speed improvements" where do I put this?

[ornate elk]

in the webui-user.bat

at the line Commandline_Args=

[median trellis]

hi there

[rigid ingot]

Beginner question - I've a Jetson Orin AGX 64GB. I've managed to get Ollama/open webui running.

I wanted to get stable diffusion running, but I can't get it past this error:

RuntimeError: CUDA error: no kernel image is available for execution on the device
Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.

Stable diffusion model failed to load
Applying attention optimization: Doggettx... done.
Exception in thread Thread-2 (load_model)
There's a larger stack trace, if that's useful. I can't figure out how to "enable device-side assertions". I'm running with python webui.py

[azure magnet]

How to fix those?

[azure magnet]

So now i should wait, right?

[next nebula]

so i am trying to install forge ai so i can work with flux loras. I keep getting this error - dubious ownership of property. the directions say to add a safe directory - git config --global etc.. presumably i do this in termimal? it doesn't specify. so i added the path as a safe directory but still getting same error

I also tried adding the git command to the run.bat file but that didn't do anything

[next nebula]

why am i blocked for asking an forge install question

[next nebula]

ok let's try another question. where do i find .gitconfig file?

never mind, found it, amended it , and good to go

[hollow ocean]

Does anyone know what might cause the Lora tab in A1111 to have the edit/delete buttons that normally pop up when you hover over the lora no longer show up? I have even done a fresh install and I don't see them there either so I am unsure what to look at next. I desperately need to sort my LORAs

[slow cedar]

why are you using a1111 😔

@sharp osprey ask here broski

[fair oxide]

4050

[ornate elk]

Okay then you should use xformers

[fair oxide]

but how

wheres the venv?

[sharp osprey]

I have trouble with loading models in stable, they’re SDXL, they never load so I can’t gen. I have a 6650xt and the memory goes up alot when I tried to load them in

[slow cedar]

THEY USE AMD BTW

oops

all caps

mocking bird has amd

[sharp osprey]

Yes

[slow cedar]

all caps lmaooo

[sharp osprey]

I would post a picture but there’s a storm outside and my internet is being wonky

And discord isn’t loading on my pc

[ornate elk]

Do you have it installed with directml or zluda as Backend?

[sharp osprey]

Directml I believe, I had zluda but deleted the folder

[ornate elk]

Edit the webui-user.bat.
At the line commandline_args=
You add --medvram --xformers
Then save and relaunch

[sharp osprey]

Then reinstalled stable diffusion again with directml

[fair oxide]

i did but all it says is no xformers found

[sharp osprey]

Xformers is only on Nvidia cards right?

[ornate elk]

Ah okay, directml is the problem.
Its not as optimised as zluda.
It uses more vram and you can't use sdxl models with it.
Zluda has less issues

[ornate elk]

Can you show me the cmd log?

[sharp osprey]

I wouldn’t know how to start getting rid of directml as a backend

I’m just very confused on all of this

[ornate elk]

You would need to delete the webui.
And then start again by using my install Guide from the pinned messages

[sharp osprey]

Pinned from where? This channel?

Oh nvm

Ty

[fair oxide]

i did that from the guide

[ornate elk]

Follow the Auto1111 with Zluda IR Forge with Zluda guide.
Let me know if you have any questions

[modest ocean]

Why do some models, even at 512x512, have no errors in the console but images generated are a grey rectangle? (Is it because I have AMD?)

[ornate elk]

Your launching it wrong.
Always and only double click the webui-user.bat to launch it

[fair oxide]

as admin or not?

[ornate elk]

How large is the checkpoint file you tried to use?

[ornate elk]

Nope not as admin

[modest ocean]

Default settings

And it usually happens on models that have 300 megabytes.

Or less

[ornate elk]

AFK, driving now (20mins)

[fair oxide]

okay, so thats why there was not any venv, now i can see venv, looks like the launch.py installed modules directly on python3.10

[ornate elk]

Models are always larger than 1gb
So 300 MB is a lora mostly

Loras are not usable standalone

[modest ocean]

So do i need model with lora, using both somehow?

[ornate elk]

Yep a lora is an addition to a model

Also what's your GPU?

[modest ocean]

5700xt

[ornate elk]

And which webui do you use?

[modest ocean]

Automatic

[ornate elk]

With directml or zluda?

[modest ocean]

Zluda

[ornate elk]

Okay then you shouldn't have issues when using 1.5 or sdxl models

2-6gb large files

[clear pawn]

Can anyone help me with getting flux gym installed locally on Mac?

[fair oxide]

Btw I had said that the SD doesnt use nvidia on optimus laptops when MuX is disabled(hybrid is enabled)

Can anything be done to this or should I just assume the devs won't be hasling about MuX things.

[fossil dew]

mmm yes, a new error.
Not sure what it means.

[ornate elk]

Seems like a browser error / connection error

Whitelist the webui in any browser adblocker

Also don't use opera vpn etc

[ornate elk]

You can let the webui use the GPU by adding --device-id 1 to the webui-user.bat

[azure magnet]

Someone could help me?

[tribal anvil]

How do I use 2 loras at once?

[limber walrus]

Hi guys, how can i solve this ? My trainings seems to have troubles with epochs

[fossil dew]

okidoki-- thank

[azure magnet]

Someone for a help

[solemn saddle]

Hey I need some help, I just had to reinstall Automatic1111 and i followed the instructions and I'm getting this error

[keen sequoia]

Hello guys, am getting kinda desperate here trying get openpose to work, but it just doesn't, i have the preprocessed image with bones, but when i generate img, the pose does not apply, help 🥲 🥲 🥲

[solemn saddle]

just as a dumb idea, try enabling low vram, when i had this issue before, for whatever reason the low vram option fixed it

[ornate elk]

Hey, your using the wrong python version

Uninstall python 3.13
Then install 3.10.11 64bit
Then delete the venv folder and relaunch the webui-user.bat

[solemn saddle]

Thank you

[keen sequoia]

Now it doesn't want to generate at all, some error pop up 🥲

[ornate elk]

What does the cmd log shows?
Make sure controlnet extension is updated and the webui too

[keen sequoia]

All is updated, i just install everything like yesterday, here's the cmd screenshot

[ornate elk]

Can you copy and paste the full cmd log?

[ornate elk]

How large is the model file you use?

[keen sequoia]

The openpose model or the one that generate the img itself?

[ornate elk]

The one for the image generation

Desiremix v4 something

[keen sequoia]

This one

[ornate elk]

Hmm okay so its sdxl/pony/illustrios based

That should work with the controlnet++ promax model

[keen sequoia]

Thats why am so confused...

[ornate elk]

The error says controlnet type unknown

Select openpose

[keen sequoia]

I did that before, doesn't work as well 😦

[solemn saddle]

@ornate elk erm what is numpy? this is entirely new never seen this before

[ornate elk]

Relaunch that error should not come again

It's a bug

[solemn saddle]

2nd launch and same error, already checked that

[ornate elk]

Wow now its using python 3.8

You need 3.10

[solemn saddle]

I installed 3.10! wtf how many python versions do i have installed

[ornate elk]

Uninstall all the pythons you have xD

[solemn saddle]

I blame you for releasing the snakes!

[ornate elk]

Then delete the venv folder again before relaunching

[ornate elk]

Have you any browser adblocker?
Make sure the webui is whitelisted in it

Then relaunch the webui again

And leave everything default

[keen sequoia]

Yap its whitelisted, i even tryed with adbloker removed from browser completely, still no luck

[ornate elk]

Have you tried a different image gen model?

Needs to be also sdxl based

[keen sequoia]

Yeah ofc, tryed this one just now, i think its sdxl... i hope? 😄

[ornate elk]

Move the controlnet model into models/controlnet

Not into the extensions folder

That was the old way

[keen sequoia]

Its in extension folder because it wasn't working when it was in models/controlnet to begin with, but i move it back to models/controlnet now, still nothing 🥲

[ornate elk]

It should work in models/controlnet
Have you tried an other browser?

For example opera and brave can block to much.
Try firefox or edge without extensions

[keen sequoia]

Gonna try firefox

Nope

[sharp osprey]

I have a problem generating on forge and it's giving me a cuda error everytime. I just freshly downloaded this to gen SDXL images. I have picture of my prompt for issue

I'm on Amd too

[keen sequoia]

Firefox didnt save the day

[ornate elk]

Okay then delete the ui-config.json and the config.json
Then relaunch the webui-user.bat

That will reset all settings

[ornate elk]

Thats why I said you should use my AMD install guides like Forge with Zluda.
Its in the first link of the pinned messages

[sharp osprey]

Ok then I’ll try that…

[ornate elk]

The normal forge works only on nvidia

So you can delete the folder

[ornate elk]

And open up a cmd and run
Pip cache purge
To start clean

[sharp osprey]

Do I download forge with zluda this time?

[ornate elk]

Read carefully and follow the steps.
Feel free to ask if anything isnt clear

[keen sequoia]

Didnt work

[ornate elk]

Can you show me the cmd log again?

Also pls show your extensions folder

[ornate elk]

Delete the two shot extension.
Its outdated since 2 years

Then delete the venv folder and relaunch

What's your GPU btw?

[keen sequoia]

4070ti super

relaunching

[keen sequoia]

Still not working

[ornate elk]

Damn

[keen sequoia]

[ornate elk]

Its very late so I need to get some sleep. But we can fix it tomorrow.

[keen sequoia]

Sure, I really appreciate it the help btw, i want it to work so bad 😭

Have a good night, and see you tomorrow, hope we can fix it

[ornate elk]

In sure we can fix it. I have the controlnet union model working so we can test and compare then

Gn

[red cobalt]

im trying to download this

but it says error code 2

[simple jetty]

I tried generating an image using SD3 with a AMD Radeon RX 7900 XT and it took over 13 minutes. What is going on? Why does it switch to 'slow mode'? The last SD was able to generate in several seconds.

[red cobalt]

how to get good generations

stable diffusion

everything looks so bad

[sterile leaf]

Same for me. I got AUTOMATIC1111 and it isn't coming out well. I am wanting to use it for making the art of a comic I am working on. I thought character consistency would be my biggest hurdle, but basic anatomy seems to be the first step here.

[red cobalt]

we need to do lora or models or smth

im not sure

[sterile leaf]

I am very new to this, so I am looking to learn (I have done some light coding before, but ai and generative art are all new to me).

[red cobalt]

same

i havent coded but

im a fast learner

or i dont forget it after i learn it once ig

[fossil dew]

Go to Settings then in the search look up K Sample and this should pop up: Enable it.

It helped me when I first started and everything was I genned was getting what I called "Rasin eyes"

also Adetailer helps a ton. Its an extension that can be added in the webui.

[sharp osprey]

Can anyone help me? I’ve download forge and zluda to use less ram for generations but whenever I try doing SDXL models, my system slows up or it doesn’t load at all. I’m on Amd, have a 6650xt, and 16gb ram

[sharp osprey]

no it I still can't load models that are sdxl... I think the amount of Vram I have is not allowing it generate, I only have 8 Vram

Yeah my forge just either doesn’t load or it crashes after a couple of minutes due to memory, I tried changing the parameters but it doesn’t do much help

I could use comfyui since SDXL models load fine for me over there just any type of stable I get the same result failed model starting slow method

[slow cedar]

Bahahaha ss your ui

@lucid yoke

Take bets euro

[lucid yoke]

Webui a1111 old guide

Spam

[slow cedar]

My guess is default settings

[lucid yoke]

Id love to get you going but I'll be home late (+12 hrs from now) if you still are stuck then just @ me

[slow cedar]

The name that's typing

Peak

[lucid yoke]

You scared them off @lucid plank what's up

[slow cedar]

😔

[red cobalt]

i needed a model lmao

[slow cedar]

If you still can't good images lmk lmao

[red cobalt]

is there a way to get 1024x1024 faster

[red cobalt]

its still not that good 1s

eyes are kinda bad

[slow cedar]

Get the adetailer extension

It'll fix the face/eyes

[red cobalt]

how

[slow cedar]

https://github.com/Bing-su/adetailer

[red cobalt]

also how do u uh

[slow cedar]

Then in a1111

Go to uhh

[red cobalt]

make it so they actually have a mouth and

[slow cedar]

Extensions tab

[red cobalt]

nose

[slow cedar]

Install from url

I think

[red cobalt]

got adetailer

[slow cedar]

I'm not looking at my ui rn