🤝｜tech-support

[azure ivy]

yes please thanks

[past birch]

where are the settings?

[naive dragon]

idk what happend but i upscaled a image and it went fully gray lol

[naive dragon]

that explains it

[ornate elk]

What settings?

You can drop the files into the library folder and overwrite

[naive dragon]

just clicked this

[naive dragon]

512x512 so it got doubled

[ornate elk]

Dont

It will use the setting of hires fix.
And if you didn't set them correctly it will error out

So for hires fix set the hires steps to 10, and denois to 0.5

Upscaler for Example to ESRGAN

[noble hinge]

So using Stable Audio, was really happy with it. I was getting the nice three minute audio generations I signed up for.
Logged in last night to find that they had changed the length, and now all the service offers is 90 second downloads.
Would not have signed up for it if I knew that was going to happen

[naive dragon]

after changing all this is there a possibility to save stuff like the settings i changed btw?

[chilly knot]

CS10 you know something aboute animatediff?

[onyx solar]

3 minute tracks is still the max duration

[ornate elk]

Maybe in the ui-config.json

[chilly knot]

cs10 i have error in animatediff like in unsupport models

its need power gpu?

[vestal dove]

🇭 🇪 🇱 🇵
(was following video: L3: Latent Upscaling in ComfyUI - Comfy Academy)
I use ComfyUI and I got a problem.

1. I generate image using pos + neg prompt using PonyDiffusion v.6 XL.
2. Then I use node Upscale Latent By (method: bislerp, 2x). And rerender image.
3. I also try to use upscaler - anime4x.

The problem is every time I use Upscale Latent By (no matter what upscale method I use) the image becomes very ugly with a lot of defects. With upscale models everything goes fine.

img1 - workspace, img2 - generated img using prompt, img3 - generated img using Upscale Latent By

[noble hinge]

Nope.

[onyx solar]

Yup

[onyx solar]

Because you have the wrong model chosen for 3 minute tracks.

[noble hinge]

which model will do it?

[noble hinge]

one moment

[noble hinge]

That did it. Thanks

[ornate elk]

Can you show me the cmd log?

[mortal compass]

Ahh no way.. That sucks. Thanks for trying it out

[chilly knot]

first i need understand my settings fine

[chilly knot]

EinopsError: Error while processing rearrange-reduction pattern "(b f) d c -> (b d) f c". Input tensor shape: torch.Size([2, 4096, 320]). Additional info: {'f': 16}. Shape mismatch, can't divide axis of length 2 in chunks of 16

if i choice batch size 16

but if i choice batch 1

as in unsupported models

[past birch]

@ornate elk hi sorry for the wait what do I need to change in the settings?

[ornate elk]

but you can upscale sdxl images with controlnet tile 1.5, but you need to select an 1.5 image for it to work

[ornate elk]

whats your gpu and whats inside your webui-user.bat?

then i also need the full error that you get in the cmd window

[brittle lantern]

6600 xt and it still uses my CPU 💀 I'm just gonna go back to full boot linux

[ornate elk]

did you followed my guide?

[brittle lantern]

Yh I did

That's how I got here

[ornate elk]

i had to update it yesterday, so better recheck

[brittle lantern]

I used it today

So it wouldn't make a diff

[ornate elk]

okay then show me your path settings

[brittle lantern]

If u updated it yesterday

[ornate elk]

and your webui-user.bat

[brittle lantern]

Alr hold up I'm downstairs eating rn

[brittle lantern]

question do i put that command line in or is sum wrong?

[ornate elk]

no dont add this

it will switch to cpu mode then

[brittle lantern]

what do i do then

[chilly knot]

gtx 1050 ti
bat file-set COMMANDLINE_ARGS=--theme=dark --xformers --medvram --no-half-vae

all like you said

[brittle lantern]

im pretty sure that was for me

[chilly knot]

why 😃

[brittle lantern]

wdym why cuz im having the problems

[chilly knot]

but he message to me

[ornate elk]

okay, you still need to show me the CMD log when you get the error

[ornate elk]

you show me your path settings

[chilly knot]

cs10 remember i try all argumensts what you give me for fix problem with model and this still not works

maybe 1050 cant works with animate diff

[ornate elk]

okay looks good, and you did download the optimised rocm library files for 6600xt? and placed them inside the library folder of rocm?

also is your driver updated?

[ornate elk]

its unlikely yes, as animatediff needs a lot of vram

[brittle lantern]

yh the file i downloaded dont have a library folder in it so i had to make one then put in the files

[ornate elk]

did you used this?

[ornate elk]

okay, and you restarted the PC after that?
also show me the full cmd log when launching the webui-user.bat

[brittle lantern]

i did

[ornate elk]

commit has none, means you didnt git cloned the repo or you moved it afterwards, also i dont see the launch arguments from the guide in it

the webui version is also outdated

the zluda fixes are missing

[brittle lantern]

i followed ur whole tutoral so idk what to do now

[ornate elk]

you need to follow it again, delete the whole stable-diffusion-webui-directml folder
then git clone the repo correctly, edit the webui-user.bat like stated in the guide and it will work
please read it carefully 🙂

[chilly knot]

but i fix many errors in facefusion maybe i can help anyone with this

[ornate elk]

let me know if you have any issues, git cloning the repo

the 1.9.3 version has fixes for zluda, thats why your 1.9.0 version didnt worked

[brittle lantern]

ive followed everything thats said

[ornate elk]

okay, whats in your webui-user.bat?

[ornate elk]

there is the error

you need --use-zluda not use-zluda

[brittle lantern]

and we back to 1

[ornate elk]

you didnt cloned it into a folder you created, that could cause permission errors

also update your python to 3.10.11 64bit

[brittle lantern]

right so i do the git command then space and put in the folder i want it to install in correct?

[ornate elk]

nope, you make a folder, click into the explorere bar, type cmd and hit enter, then you just paste the whole command in and hit enter

[brittle lantern]

i got it

[ornate elk]

okay, so now everything is correct. but now make sure your gpu driver is at least on Adrenalin 24.1.1

also show me your rocm/bin folder where the librarys are

[brittle lantern]

its on the latest driver

[brittle lantern]

if u mean whats in it

[ornate elk]

oh sry in rocblas

[brittle lantern]

i had to make the library folder as the file for 6600 xt didnt have the library folder ur tut mentioned

but it had the files

[ornate elk]

okay to make sure you have the GFX1032 and not gfx1031 files andmove them into the library folder again, then restart the PC

[brittle lantern]

done now whats next

[ornate elk]

now relaunch the webui-user.bat

[brittle lantern]

i figured i did that and ive still got the same issue

so ive tried basically almost everything

just to go back to square one

[ornate elk]

can you go into the library folder

and check if there are gfx1031 files?

[brittle lantern]

1032 only

i downloaded the 6600 xt opti files linked which has 1032

6700 has 1031

do i download that one then if i need it?

[ornate elk]

okay thats good

1032 is for 6600xt

[brittle lantern]

got u now what then whats the issue

[ornate elk]

now its hard to guess

what you should do is to open up a cmd as admin and run
sfc /scannow
that will detect and repair corrupt OS files.
Then i would reinstall the ROCm HIP SDK and then do the library steps again + restart again.

[brittle lantern]

doesnt work same error

yk atp i give up thx for the help tho man @ornate elk

[ornate elk]

sry that we couldnt find the cause of the error :/ it works for the most people but every system is different so we cant know what causes it.

[timid notch]

i think i find a bug, the whole process will stop when everytime i add this node,~~check the video and you can see this happen~

[brittle lantern]

It's alr I'll just buy a nvidia GPU and sacrifice the 7900 xt gaming performance I was gonna get for £600

[ornate elk]

@brittle lantern last thing you could try,
open up a cmd again and type
pip cache purge
then relauncht the webui-user.bat

[brittle lantern]

nope

[ornate elk]

okay, again a thing you can try:
edit the webui-user.bat
in and empty row add:
set HSA_OVERRIDE_GFX_VERSION=10.3.0
then save and relaunch

[brittle lantern]

nope

Either my pc is clapped or sum is wrong

I'll just get a nvidia gpu for £600 instead of the 7900 xt and sacrifice gaming performance

Ty anyways

[warped canyon]

rich boy. popping gpus left and right

[brittle lantern]

It's for my b day I usually can't get gpus 😅

[warped canyon]

aight

[ornate elk]

A 7900xt is a beast card when you compare it to Nvidia cards at the same price range

It would also work better with zluda as the rocm librarys don't need to be changed

[brittle lantern]

Ik I'm well aware I'm a AMD fan but I cba dealing with these issues

[ornate elk]

Yea its unlucky :/

[brittle lantern]

What's the best option 4070 or 4070 super or do I go used or refurb 3090?

[ornate elk]

3090 would be great, but a 4070ti beats it in performance (gaming)

For Ai stuff a 3090 is great

[brittle lantern]

I play pretty much esports games at 1080p I think both is overkill for that 😅

[ornate elk]

True xD

[brittle lantern]

So ig 3090 it is

I can get it for £240 GBP if I do trade ins which I can

My old cpu my 1060 6gb and then my current 6600 xt as I'm getting a 3090

[noble belfry]

I dont know your tech knowhow level wyn, do you know how to physically install this, have your system accept it and change to a higher PSU? You're absolutely capable of it i'm sure, but you need to check things like your motherboard, space to rack it inside, cooling etc.

[brittle lantern]

Dw I build pcs and help with support for most of my job yes it's fine it works I'm assuming it also could be a windows install issue

[noble belfry]

yeah sorry if i sounded patronising. tech problems are good reference to look back on for others.

[sinful needle]

hey can anyone tell me how to make it so animatediff doesnt straight up try to animate literally everything in a scene? its kinda ridiculous.

[wicked leaf]

Hello everyone, I am trying to create a RAG pipeline using haystack-pinecone and hugging face, but for some reason i cannot upload my vector-embeddings on pinecone, can somebody help me out? Its a very small code and won't take much time.

[azure ivy]

hey is it possible to send it to me tnx

[ornate elk]

https://github.com/Scholar01/sd-webui-mov2mov/issues/146#issuecomment-2054157728

[azure ivy]

thank you

[sinful needle]

😦

[pastel coral]

not sure what to do

this also happens when i follow what the github says to try and fix this so im stumped i guess

[pastel coral]

nvm we ight

[vocal fox]

do you think anyone can help me with this problem:

So I was using SDXL and it was slow as shit, so I decided to follow these reddit instructions (here: https://www.reddit.com/r/StableDiffusion/comments/168sad2/comment/jyy63za/?context=3&share_id=0BoI2GicEf26WJP_rmU3i&utm_content=1&utm_medium=ios_app&utm_name=ioscss&utm_source=share&utm_term=1) and I broke everything...

It messes up on step 2 with the error message:

Any ideas what I can do to fix this?

[frank wasp]

hye

cananyone help mepls

ERROR: Could not find a version that satisfies the requirement torch==2.1.2 (from versions: 2.2.0, 2.2.0+cu121, 2.2.1, 2.2.1+cu121, 2.2.2, 2.2.2+cu121, 2.3.0, 2.3.0+cu121)
ERROR: No matching distribution found for torch==2.1.2

i keep geetting this error

[solemn moon]

my go-to is to delete venv folder, then rerun .bat file
@vocal fox @frank wasp
not sure if it will help but worth the try

ya'll should also post yr command line args and state yr setup for people to try assess better

[sudden hawk]

hello,
i dont know what happens with my SD since i add embedding > when i try to generate: the rendering is a grey screen while I can see the pic changing during the chargement. Even when I go in output the images are fully grey. does I have to add arguments ? and which one ?

[karmic crown]

do not use python 3.12, use 3.10

[ornate elk]

Your third party antivirus Programm broke the Python installation

[ornate elk]

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

[humble cobalt]

hello, idk whats happening when im generating a second image, at the end of it it just black my screen for a moment and then the cmd shutdown and it shows me this error (even with pagefile modified) :

minidump-analyzer.exe – Application Error
The Breakpoint Exception
A breakpoint has been reached.
(0x80000003) occurred in the application at
location 0x00007FFB88E1695E.
Click OK to terminate the program.
Click Cancel to debug the program
Cancel

and that :

[ornate elk]

Hey, go into the stable-diffusion-webui folder.
Click in the explorer bar (not searchbar)
There type cmd and hit enter.
Then type
Git pull
And hit enter, Then type
Pip cache purge
And hit enter.
Then relaunch the webui-user.bat

[ornate elk]

Hey, What's your GPU? What's in your webui-user.bat?
Did you used sdxl or 1.5 model?

[humble cobalt]

i have rtx 3060 ti. i use --xformers --medvram-sdxl --no-half-vae with a Pony XL merge model. It was working at the beginning but it started doing these error out of nowhere i tried to relaunch my pc and still nothing

[ornate elk]

Okay, how much RAM do you have?

[humble cobalt]

16, with 16000-24000 pagefile

[vocal fox]

thanks, but i just decided to install forge, which actually was painless to install somehow

[ornate elk]

Oh okay

[ornate elk]

And you restarted the PC after modifieng the pagefile?

[humble cobalt]

yeah, i modified the pagefile yesterday

[ornate elk]

Okay, also make sure the webui is whitelisted in any browser adblocker

Also make sure Wallpaper engine isnt running

[humble cobalt]

but i think my windows is doing shit, like my task bar is glitched, icons are overlaping

[humble cobalt]

i dont have adblocker on firefox and wallpaper engine

and my windows is updated. I know windows 11 is not the best optimized, it can be the problem ?

[ornate elk]

Okay then open up a cmd as admin and run
SFC /scannow
That will check and repair corrupt system files

[sudden hawk]

i have rtx nvidia 3070 and in my webui : --xformers --no-half-vae --autolaunch --medvram-sdxl

[humble cobalt]

do i need to restart the pc ? it detected corrupted files and repaired it

[ornate elk]

Yep

[ornate elk]

Okay make sure the embedding is compatible with the model version

[sudden hawk]

ok, I have delete the embedding from the folder but I still have grey images

[ornate elk]

Then make sure the model you use is a model and not a lora

Models are always larger than 2GB

[sudden hawk]

yes I know, I use Fusiondanceanime as checkpoint and I know it's a checkpoint cause it was working before I add embedding...

[ornate elk]

Okay good

And you restarted the webui?

[sudden hawk]

yes, I think the problem comes from Fusiondanceanime cause when i generate with another checkpoint it works.. I'm gonna delete and make a new dl for this checkpoint

Thank you for reply and your help @ornate elk

[ornate elk]

Np 🙂

[frank wasp]

i do but it keeps selecter ing pthron 3.12

selecting

some how i cant select pyhton 3.10.6

[humble cobalt]

I restarted it but still getting an error, but this time its "WerFault.exe - Application error
The application failed to start correctly
(Oxc000012d). Click OK to close the application.

The instruction to 0x00007FFBD8B95474 uses the address
memory 0x00000000000000000. The state of the memory cannot be read.
Click OK to complete the program.
Click Cancel to debug the program"

[ornate elk]

You need to uninstall python 3.12
Then install python 3.10.11 64bit
Then delete the venv folder and relaunch the webui-user.bat

[humble cobalt]

Same error when i didnt modified my pagefile

But i did now, and idk why its still doing this error

[ornate elk]

Do other models work?

[humble cobalt]

Doesnt seem to

I'll restart my pc and try a different model to see

[frank wasp]

thanks alot

[frank wasp]

it works but imgettng another errror now WARNING: There was an error checking the latest version of pip.
Traceback (most recent call last):
File "C:\AI\stable-diffusion-webui\launch.py", line 48, in <module>
main()
File "C:\AI\stable-diffusion-webui\launch.py", line 39, in main
prepare_environment()
File "C:\AI\stable-diffusion-webui\modules\launch_utils.py", line 386, in prepare_environment
raise RuntimeError(
RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check

[ornate elk]

What's your GPU?

[frank wasp]

amd ryzen

[ornate elk]

Not CPU

[frank wasp]

and radeaon 6900

xt

[frank wasp]

Amd Radeon 6900xt

[ornate elk]

Okay, then you need a different webui version.
I made a guide for AMD cards.
Here: #🤝｜tech-support message

You can delete the stable-diffusion-webui folder before starting with the guide

If you have any questions, feel free to ask me 🙂 I'm also on AMD

[humble cobalt]

With an other model i can generate 2 images. After that there is no black screen or error windows message but just "connection errored out" and the cmd showing "press any key to continue"

[ornate elk]

Is the other model and sdxl one too?

What is the resolution?

Did you used upscaling?

[humble cobalt]

The other model is a sdxl too yes. The resolution is 856 x 1192 without upscaling

[ornate elk]

And you used batch size 1 and batch count 1 ?

[humble cobalt]

Yep, clip skip 2, 20 steps with Euler A

[ornate elk]

Okay, then check the RAM usage in taskmanager when generating

And make sure you have at least 15gb free space on C

[humble cobalt]

Im already at 3.5 used

With Firefox and windows system stuff

[humble cobalt]

So i have like 12 GB free

[ornate elk]

Then make more free space

If the pagefile can't dynamicly increase its size, it will error out

[humble cobalt]

Oh wait im dumb i though you was talking about RAM mb

I have 35 GB free in my local disk

When i launch the webui, without generating anything i am at 4.7 GB RAM

[ornate elk]

Is your webui updated? And the graphic cards driver too?

[humble cobalt]

Remaining

[humble cobalt]

1.9.3

And yes the driver is updated

[frank wasp]

hey i tried everything but one stepim not sure

[ornate elk]

Okay, which step?

[frank wasp]

adding o path

i put my zleda folder outside of the zleda- i made

[ornate elk]

You need to have a C:/Zluda folder or wherever you placed it

And that goes into the Path

Not the SD-Zluda folder with the webui in it

[frank wasp]

i did

and what do i add as variable?

like this? @ornate elk

[ornate elk]

No not like that

Remove it

[frank wasp]

ok i did

[ornate elk]

Your at the User variables.
But you need to look below at the System variables.
There click on PATH, then on edit,
then on New
Then add the path to the folder.
And the %hip thing as second

[frank wasp]

like this

[ornate elk]

No

Please revert it

Click on PATH then on edit

[humble cobalt]

i tried and it showed me an error on the webui "OutOfMemoryError: CUDA out of memory. Tried to allocate 20.00 MiB. GPU 0 has a total capacty of 8.00 GiB of which 2.54 GiB is free. Of the allocated memory 4.16 GiB is allocated by PyTorch, and 191.74 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF"

[ornate elk]

Make sure you always launch the webui-user.bat

Then you also should get the tiled vae extension

[humble cobalt]

i did, i could generate a few images but after when i clicked again on generate it instantly showed this

[frank wasp]

and then

[humble cobalt]

its saying "try setting max_split_size_mb" but idk what it is

[ornate elk]

Now on New

Then add the path to the zluda folder

And the hip path

[ornate elk]

Nah that won't help

Try the tiled vae extension

[humble cobalt]

okay, btw if it can help, after the error when i wanted to load a page on my browser it was saying i didnt have enough memory left to load it*

[ornate elk]

And now the %HIP path from the guide

[frank wasp]

last step

wich do i select

[humble cobalt]

tried tiled vae option of the extension and still not

[ornate elk]

Why did you removed the zluda path?

[ornate elk]

Okay then lower the resolution

[frank wasp]

i didnt

[ornate elk]

But its missing ?

[humble cobalt]

but idk why its doing that. Before it was working perfectly

[ornate elk]

In your top screenshot its there and then its not

[frank wasp]

i added it c=agaiab it went on accident i guesss

[humble cobalt]

maybe i should try to reinstall it again

[frank wasp]

i put it back

but now im stuck

the tutorial says i have to be here to add the %

but i cant ffigure out where

how do i ge this?

%HIP_PATH%bin

@ornate elk

[ornate elk]

Click on new

And add the %hip path line

[frank wasp]

ID DID IT

[ornate elk]

Perfect

[frank wasp]

i tought this would have to eed ath too so iwas spaicng

path*

and nnow just start web ui user

?

[ornate elk]

Yep

[frank wasp]

https://tenor.com/view/roger-gol-d-roger-king-of-pirates-laughing-one-piece-gif-26533390

[willow kettle]

Hey @ornate elk I know you mostly deal with people having trouble here so I just wanted to mention a success story. Last night I followed your zluda guide for AMD cards and it worked great for me on the first try. Thanks!!!!

[ornate elk]

Hey, glad it worked 🙂 thanks for letting me know

[frank wasp]

launch.py: error: unrecognized arguments: --use-zluda --skip-ort

i ahve thisnow

[ornate elk]

Okay, show me the webui-user.bat

[frank wasp]

@ornate elk

[ornate elk]

Add the arguments again. Remove the space before --use-zluda

If you get the error again, show me a screenshot of the cmd

[oak bough]

Does it need to be created if we're adding models to StableSwarm-UI?

[frank wasp]

when i delet the --use-zluda --skip-ort it boots now

but it keeps remembering it with the -- (space)

@ornate elk

[ornate elk]

show me the full cmd log

[frank wasp]

this one above?

[ornate elk]

oh

oh wait

your in the wrong folder

delete the AI folder

its from your OLD install

then go into the SD-Zluda folder

there ist the new webui

there edit the webui-user.bat

and then launch it

[frank wasp]

oh shit

thats why

venv "C:\SD-Zluda\stable-diffusion-webui-directml\venv\Scripts\Python.exe"
WARNING: ZLUDA works best with SD.Next. Please consider migrating to SD.Next.
Failed to load ZLUDA: Could not find %HIP_PATH%. Please install AMD HIP SDK.
Using CPU-only torch
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-6-ga92b859e
Commit hash: a92b859e4090a1d33576036a688a44bb6b5d55b7
Traceback (most recent call last):
File "C:\SD-Zluda\stable-diffusion-webui-directml\launch.py", line 48, in <module>
main()
File "C:\SD-Zluda\stable-diffusion-webui-directml\launch.py", line 39, in main
prepare_environment()
File "C:\SD-Zluda\stable-diffusion-webui-directml\modules\launch_utils.py", line 592, in prepare_environment
raise RuntimeError(
RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check
Press any key to continue . . .

do i need to make te aths again

[ornate elk]

Did you installed HIP SDK ?

[frank wasp]

i did

but mab o nthe old

hip sdk

i will see

[ornate elk]

Yea recheck the guide steps

[frank wasp]

i did

[ornate elk]

Also you need to restart the PC after installing HIP

[frank wasp]

ah ok

i didnt do that

i will try

i restarted

same error

[idle sapphire]

i tried downloading it with python 3.10.11 i put -lowvram as i have 4gb vram in my gpu ( i use integrated ryzen 5600g graphics)

this is the full output

i put this on the .bat

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--use-directml --skip-ort --lowvram --opt-sub-quad-attention --opt-split-attention --no-half-vae --upcast-sampling

call webui.bat

[frank wasp]

anyone know what can try? venv "C:\SD-Zluda\stable-diffusion-webui-directml\venv\Scripts\Python.exe"
WARNING: ZLUDA works best with SD.Next. Please consider migrating to SD.Next.
Using ZLUDA in C:\SD-Zluda\stable-diffusion-webui-directml.zluda
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-6-ga92b859e
Commit hash: a92b859e4090a1d33576036a688a44bb6b5d55b7
Traceback (most recent call last):
File "C:\SD-Zluda\stable-diffusion-webui-directml\launch.py", line 48, in <module>
main()
File "C:\SD-Zluda\stable-diffusion-webui-directml\launch.py", line 39, in main
prepare_environment()
File "C:\SD-Zluda\stable-diffusion-webui-directml\modules\launch_utils.py", line 592, in prepare_environment
raise RuntimeError(
RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check
Press any key to continue . . .

[ornate elk]

You didn't followed my guide.
You installed it the wrong way by downloading it as zip file.
Then the next error is that you placed it on desktop with OneDrive sync enabled

[ornate elk]

Can you show me the environment path and the zluda folder again?

Can look later at it, I'm afk

[frank wasp]

cool i thik i funnd it vev folder

deleteing now

venv "C:\SD-Zluda\stable-diffusion-webui-directml\venv\Scripts\Python.exe"
WARNING: ZLUDA works best with SD.Next. Please consider migrating to SD.Next.
Using ZLUDA in C:\SD-Zluda\stable-diffusion-webui-directml.zluda
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-6-ga92b859e
Commit hash: a92b859e4090a1d33576036a688a44bb6b5d55b7
Skipping onnxruntime installation.

| You are not up to date with the most recent release. |

Consider running git pull to update.
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
C:\SD-Zluda\stable-diffusion-webui-directml\venv\lib\site-packages\pytorch_lightning\utilities\distributed.py:258: LightningDeprecationWarning: pytorch_lightning.utilities.distributed.rank_zero_only has been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it from pytorch_lightning.utilities instead.
rank_zero_deprecation(
Launching Web UI with arguments: --use-zluda --medvram-sdxl --update-check --skip-ort

[ornate elk]

Looks good

[frank wasp]

its better than befoe

now i need to update i guess but update what xd

[ornate elk]

You can update it later when it works

By putting git pull into the webui-user.bat

[frank wasp]

it doesnt oot

i did

[ornate elk]

What does the cmd shows?

[frank wasp]

venv "C:\SD-Zluda\stable-diffusion-webui-directml\venv\Scripts\Python.exe"
WARNING: ZLUDA works best with SD.Next. Please consider migrating to SD.Next.
Using ZLUDA in C:\SD-Zluda\stable-diffusion-webui-directml.zluda
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-7-g216340dc
Commit hash: 216340dc8778c1ce4d19ee93621258d7986015f5
Skipping onnxruntime installation.
You are up to date with the most recent release.
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
C:\SD-Zluda\stable-diffusion-webui-directml\venv\lib\site-packages\pytorch_lightning\utilities\distributed.py:258: LightningDeprecationWarning: pytorch_lightning.utilities.distributed.rank_zero_only has been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it from pytorch_lightning.utilities instead.
rank_zero_deprecation(
Launching Web UI with arguments: --use-zluda --medvram-sdxl --update-check --skip-ort

[ornate elk]

Let it load

[frank wasp]

ok

[ornate elk]

It can take a long time

[frank wasp]

ok im just happy if it works xd

thanksk alot btw

im trying for like 6 hours xd

[ornate elk]

Yea I can imagine the struggle xD

Thats why I made that guide for us AMD users ^^

[frank wasp]

our he best man

1year ago you helpedme out also

you hsve real patince xd

[ornate elk]

@idle sapphire let me know if you have any issue with the installation if the Directml guide:
#🤝｜tech-support message

[frank wasp]

venv "C:\SD-Zluda\stable-diffusion-webui-directml\venv\Scripts\Python.exe"
WARNING: ZLUDA works best with SD.Next. Please consider migrating to SD.Next.
Using ZLUDA in C:\SD-Zluda\stable-diffusion-webui-directml.zluda
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-7-g216340dc
Commit hash: 216340dc8778c1ce4d19ee93621258d7986015f5
Skipping onnxruntime installation.
You are up to date with the most recent release.
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
C:\SD-Zluda\stable-diffusion-webui-directml\venv\lib\site-packages\pytorch_lightning\utilities\distributed.py:258: LightningDeprecationWarning: pytorch_lightning.utilities.distributed.rank_zero_only has been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it from pytorch_lightning.utilities instead.
rank_zero_deprecation(
Launching Web UI with arguments: --use-zluda --medvram-sdxl --update-check --skip-ort

[ornate elk]

Oh wow, I didn't knew ^^

[ornate elk]

All fine, let it load (15-40 minutes)

[frank wasp]

rhipMemUnmap() + 0x85B7 byte(s)
0x00007FFB482A7A84, C:\windows\SYSTEM32\amdhip64.dll(0x00007FFB47F30000) + 0x377A84 byte(s), hipMemUnmap() + 0x62FB4 byte(s)
0x00007FFB482B2050, C:\windows\SYSTEM32\amdhip64.dll(0x00007FFB47F30000) + 0x382050 byte(s), hipMemUnmap() + 0x6D580 byte(s)
0x00007FFB482AA81B, C:\windows\SYSTEM32\amdhip64.dll(0x00007FFB47F30000) + 0x37A81B byte(s), hipMemUnmap() + 0x65D4B byte(s)
0x00007FFB482A9FB5, C:\windows\SYSTEM32\amdhip64.dll(0x00007FFB47F30000) + 0x379FB5 byte(s), hipMemUnmap() + 0x654E5 byte(s)
0x00007FFB482740B2, C:\windows\SYSTEM32\amdhip64.dll(0x00007FFB47F30000) + 0x3440B2 byte(s), hipMemUnmap() + 0x2F5E2 byte(s)
0x00007FFB482741A1, C:\windows\SYSTEM32\amdhip64.dll(0x00007FFB47F30000) + 0x3441A1 byte(s), hipMemUnmap() + 0x2F6D1 byte(s)
0x00007FFB48247357, C:\windows\SYSTEM32\amdhip64.dll(0x00007FFB47F30000) + 0x317357 byte(s), hipMemUnmap() + 0x2887 byte(s)
0x00007FFB48250F5F, C:\windows\SYSTEM32\amdhip64.dll(0x00007FFB47F30000) + 0x320F5F byte(s), hipMemUnmap() + 0xC48F byte(s)
0x00007FFBD879257D, C:\windows\System32\KERNEL32.DLL(0x00007FFBD8780000) + 0x1257D byte(s), BaseThreadInitThunk() + 0x1D byte(s)
0x00007FFBDA04AA48, C:\windows\SYSTEM32\ntdll.dll(0x00007FFBD9FF0000) + 0x5AA48 byte(s), RtlUserThreadStart() + 0x28 byte(s)
Terminate batch job (Y/N)? cc

just keep it like this?

[ornate elk]

Oh thats an erroe

Make sure your driver is updated

[frank wasp]

ok i justt didadriver upgrade and now im awiting agai

venv "C:\SD-Zluda\stable-diffusion-webui-directml\venv\Scripts\Python.exe"
WARNING: ZLUDA works best with SD.Next. Please consider migrating to SD.Next.
Using ZLUDA in C:\SD-Zluda\stable-diffusion-webui-directml.zluda
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-7-g216340dc
Commit hash: 216340dc8778c1ce4d19ee93621258d7986015f5
Skipping onnxruntime installation.

| You are not up to date with the most recent release. |

Consider running git pull to update.
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
C:\SD-Zluda\stable-diffusion-webui-directml\venv\lib\site-packages\pytorch_lightning\utilities\distributed.py:258: LightningDeprecationWarning: pytorch_lightning.utilities.distributed.rank_zero_only has been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it from pytorch_lightning.utilities instead.
rank_zero_deprecation(
Launching Web UI with arguments: --use-zluda --medvram-sdxl --update-check --skip-ort

[ornate elk]

Okay good. Make sure you don't run anything heavy like a Game or Wallpaper engine until its done

[frank wasp]

ok

this is ok right?

| You are not up to date with the most recent release. |

Consider running git pull to update.
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
C:\SD-Zluda\stable-diffusion-webui-directml\venv\lib\site-packages\pytorch_lightning\utilities\distributed.py:258: LightningDeprecationWarning: pytorch_lightning.utilities.distributed.rank_zero_only has been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it from pytorch_lightning.utilities instead.
rank_zero_deprecation(
Launching Web UI with arguments: --use-zluda --medvram-sdxl --update-check --skip-ort
Traceback (most recent call last):
File "C:\SD-Zluda\stable-diffusion-webui-directml\launch.py", line 48, in <module>

main()

File "C:\SD-Zluda\stable-diffusion-webui-directml\launch.py", line 44, in main
start()
File "C:\SD-Zluda\stable-diffusion-webui-directml\modules\launch_utils.py", line 692, in start
import webui
File "C:\SD-Zluda\stable-diffusion-webui-directml\webui.py", line 13, in <module>
initialize.imports()
File "C:\SD-Zluda\stable-diffusion-webui-directml\modules\initialize.py", line 36, in imports
shared_init.initialize()
File "C:\SD-Zluda\stable-diffusion-webui-directml\modules\shared_init.py", line 66, in initialize
initialize_zluda()
File "C:\SD-Zluda\stable-diffusion-webui-directml\modules\zluda.py", line 42, in initialize_zluda
if not test(device):
File "C:\SD-Zluda\stable-diffusion-webui-directml\modules\zluda.py", line 26, in test
return out.sum().is_nonzero()
KeyboardInterrupt
^CTerminate batch job (Y/N)?

or an error again

@ornate elk

[frank wasp]

it worked

anyone now how i genereate?

when i do i get nodthing

[trail canopy]

hey, im using automatic1111 webui and rx580 4gb gpu, and when im trying to generate image bigger than 128x128 in the last 10% of the generating image suddenly my computer turns off, i thought its power supply problem but its probably not, because when im using amd adrenaline gpu test my gpu usage goes to 100% and pc doesnt turn off but in the stable diffusion my gpu usage doesnt even go to 80% so thats weird and i also tried to limit watts usage to only 70% but didnt work, anyway im using --use-directml --skip-ort --lowvram --opt-sub-quad-attention --opt-split-attention --no-half-vae --upcast-sampling --autolaunch --no-half arguments. what should i do?

[trail canopy]

Noteworthy: The first image generation can take from 15 up to 40 minutes. First time only.

[frank wasp]

is that also with aloading scen

bc mine is nt even oading

just leave it like this then

[frank wasp]

ill try that

[scarlet sinew]

Hi Guyz i am new here can any body help me to prevent text generation on the image from stable diuffusion

tell me if can we finetune the Sdxl for not producing the text on any image, becasue stable diffusion is very bad in producing text so can we somehow stop sdxl to product any kind of text on the generated image, Currently i am using negative prompt

NEGATIVE PROMT = "text, fonts,words,3d, cartoon, anime, (deformed eyes, nose, ears, nose), bad quality,bad anatomy, ugly"

but it does not listen to the negatrive prompt so well
i want to generate canva template for christmas halloween etc with no text on it but it always put text with wrong spelling

[idle sapphire]

what did i installed as zip?

[plush cliff]

hi, I`m using A1111 webui. RTX 3050 8GB/i5 10400f/16GB RAM. so I can't enable xformers, then I use the launcher and enable xformers, it turns on but it shows as infinite loading, how can I fix this?

[idle sapphire]

i fixed it, now it works.

[plush cliff]

oh you can`t believe me) when I wrote this massage it fixed

Sorry for disturbing

[ornate elk]

First image will take 15-40 mins then it should work normaly

[ornate elk]

Make sure you use a 2gb model

Like Dreamshaper v8

The default 1.5 model is to big

[ornate elk]

Yes , just let it load

[frank wasp]

its stays half hour like this

@ornate elk i dont think its gonna work

what do ou sugest i try

[mortal compass]

I might be stupid but I fucking cannot find any info on what THESE are? I take it it's different lora's a specific user has made? ...Or? 😬

[simple roost]

I've been using XL models lately, and when I try to use any models based on 1.4, 1.5, etc, they all look good in the preview, but the final image comes out like this. Any idea what causes that?

[onyx solar]

VAE

[simple roost]

Wonderful! I had forgotten i change it to an XL VAE. Thank you for that!

[fierce hollow]

Photos are generated for a long time, I have 3070ti, everything was fine earlier, then a Windows error occurred, the computer went into reboot, now everything is generating for a very long time. Can you tell me what this might be related to?

[ornate elk]

Click generate again

TThen check the cmd

[frank wasp]

it wored

worked

thanks alot

[ornate elk]

These are Lora's for pony models

[ornate elk]

Perfect 🙂 no problem

[ornate elk]

Open up a cmd as admin and type
SFC /scannow

That will repair corrupt system files

Then make a PC restart

[fierce hollow]

thanks

[hybrid thistle]

@frank wasp oye como puedo usar los comandos para crear imágenes y videos

[ornate elk]

You can't generate images in this discord.
The bots are offline

[trail canopy]

same problem with dreamcatxherv8

it goes up to 90% ans then pc turns off

[ornate elk]

Is your GPU connected with 2 power cables or one?

Or does it have a 2 or 1 connectors?

[trail canopy]

it has only one

sapphire rx580 4gb

[ornate elk]

Hmm if the PC turns of that means mostly the power supply can't handle the power draw

[ornate elk]

What's in your webui-user.bat?

[prime wing]

hi im installing ComfyUI on Linux
there's any good reason to install it with a virtual environment (python3-venv)?

[trail canopy]

that was my guess but even with -30% power supply limit in amd gpu driver settings it does same thing, whatever im gonna use cloud or smth, anyway thanks for trying to help

[ornate elk]

Okay, no problem

[trail canopy]

do you know whats the cheapest cloud for stable diffusion with decent generating time?

[ornate elk]

Hmm idk, but check services like: runpod.io, paperspace, vast.ai

Or checkout leonardo.ai for a online webui without anything to setup

[plush cliff]

Thx, but i‘ve already solved the problem

[karmic crown]

pip install torch torchvision torchaudio on Intel Macs installs 2.2.2, and if you try to install 2.3.0 with pip install torch==2.3.0 torchvision==0.18.0, you will get an error

i just tested

it's time for a PR to a1111 repo

[willow kettle]

Does ReActor only work with nvidia?

[bright olive]

Question: I've installed Comfy first, and now i've installed Automatic1111. how do I get A1111 to recognize Comfy's models. I see readme's and tutorials going to other way, but I don't want to re download or move any models from Comfy

[noble belfry]

look up sym links 🙂

[raven goblet]

What does the "count_X, count_y_up, count_z_up, ..." syntax do in a SD text prompt? I tried asking ChatGPT, but it just told me it was a variable counter. I figured that out on my own. I see it a lot, and it seems to relate to Loras somehow, but I cannot figure out what it is for, and when it is appropriate to use.

[ornate elk]

On GPU mode yes, You have to click on the reactor settings, there switch to CPU mode.

Then it works

[sinful needle]

Anyone know how to make it so animatediff doesn't indiscriminately randomly animate ALL things in a scene? Here I just wanted the person to move, not the streetlamps lol

[weary dagger]

I might be stupid but I fucking cannot find any info on what THESE are? I take it it's different lora's a specific user has made? ...Or

||I might be stupid but I fucking cannot find any info on what THESE are? I take it it's different lora's a specific user has made? ...Or||

[ornate elk]

These?

[grim iron]

I keep getting the following error while using inpaint, I'm using a 3080 RuntimeError: Sizes of tensors must match except in dimension 1. Expected size 1 but got size 2 for tensor number 1 in the list.

[ornate elk]

Make sure you set the inpatient area to mask only, then set a desire resolution.

[frosty frigate]

Hi I am trying to set up regional prompting using the Regional Prompter extension but cannot get it to work - I keep getting an error TypeError: 'NoneType' object is not iterable
Does anyone know what this might mean?

[ornate elk]

Hey, Can you show your regional prompt settings?

[wanton plover]

hey guys help me, i have generates this with any model

How to fix it

[vocal burrow]

need more info, what settings did you use, what tool are you using (a1111, comfyui, etc), what does the log says ?

[wanton plover]

where's the log?

I'm just new to sb.

[vocal burrow]

it should be in the console

[wanton plover]

.

[vocal burrow]

ok then what settings did you use (screenshot your webpage) ? what s your hardware ?

[wanton plover]

I downloaded it yesterday, if anything I downloaded it before and everything was working fine but now it's not working well.

[wanton plover]

my pc, 1050ti, i5 7400

[ornate elk]

Please no nsfw stuff here.
Yiffimix models are badly trained. Some of them require additional config (.yaml) files. So you need that too
Newer variants don't require such file

[wanton plover]

ok

[willow kettle]

Thanks!!

[late tree]

If by virtual environment you mean like VirtualBox or VirtualMachine or something like those, where you run a different operating system through a program then you don't want to do that as the virtual machines usually don't have access to the components like GPU.

As for venv, I forget the exact reason but I believe using venv is supposed to avoid conflicts or something like that.

[open fox]

I usually install more complex applications (many requirements packages) in virtual environments. The overhead is small and I can make sure to keep the dependencies clear. If you only use comfy and do not want to install llm later or whatever in python you do not need a virtual environment

[naive dragon]

i get AttributeError: 'NoneType' object has no attribute 'lowvram' when trying to use img2img i only use --use-zluda --medvram-sdxl --update-check --skip-ort in webui-user.bat or is there something else i might be doing wrong?

[ornate elk]

What's your webui version? (At the top of the cmd)

Also what are the img2img settings?

[naive dragon]

it worked before thats why im confused 😅

[ornate elk]

Do you just try to generate in img2img?

No inpainting or upscaling?

[stray comet]

How would someone go to create a unique real life look person? Should I train a Lora on 2 different faces and hope they will blend together into a unique one when i run it after? Or maybe just generate someone random, inpaint, generate on grids and then train the lora on that? Or is there another better way?

[naive dragon]

yes mostly upscaling

[ornate elk]

And what are the upscale settings?

And sdxl or 1.5 ?

[weak mason]

Hi there! 👋🏻 I'm having some issues with the image-to-image mode in the new sd3 api from Stability AI.

When the weight gets above .8 - .9 ish it starts generating black edges on the images. See the attached examples.

Has anyone else experienced this or know if there's a fix? Honestly this seems like a bug in the API most likely

[naive dragon]

sdxl

[naive dragon]

it crashed and i dont exactly know the upscale settings i used ;/

[ornate elk]

When you upscale in img2img do that with the SD upscale script at the bottom.
Then select a square resolution (for sdxl 1024x1024) and lower the denois below 0.3
Then also set your fav upscaler at the bottom

[naive dragon]

i used denois 0.5

[naive dragon]

alright thank you

[vocal burrow]

That might be a bit much. The higher the value, the more the output will differ from the input

[naive dragon]

😅 i guess so haha

[light geyser]

can anyone help? how do i get those nodes?

i never used comfyui

got this off civitai

[river chasm]

have a issue with my stable diffusion installation

RuntimeError: Torch is not able to use GPU;

i get this
what do i do

i have a 2gb vram gpu..
its nvidia quadro k2000

[vocal burrow]

k2000 was released about 15 years ago. It s probably too old and unsupported by pytorch.

[river chasm]

so theres no way for me ?

[vocal burrow]

you re stuck using "cpu only" mode but if your cpu is that old too...... it will probably take hours to generate a single picture

[river chasm]

ahh yes

[vocal burrow]

maybe less if using lighting/turbo models. But even then it will be very slow.

[river chasm]

so theres no way to use my gpu right ?

[vocal burrow]

yup.
Ýou might want to try to force an older pytorch version that is compatible with your gpu.... But then it will probably be too old to work with SD/Auto1111. For such a "weak" gpu, it s not worth the effort honestly.

[valid mulch]

Anyone noticed a difference between using SD in one web browser vs. another? I'm pondering whether doing things in Edge vs. Chrome might free up some of my limited GPU...

[vocal burrow]

Your browser should have no impact whatsoever on the output. Can you give more details ?

[valid mulch]

Not on the output per se - just on performance. It occured to me that Chrome itself might be sucking up some of my RAM/VRAM and that using a different browser might leave more room for SD to work with.

[vocal burrow]

yes, different browser use ram/vram differently. Depending of the webengine, the ammount of tabs, extensions, etc

[cosmic furnace]

When on zluda vram also eventually clog up ? 🤨

Randomly kinda. Become like very slow if u do higher res . Until reboot sd -_-

[warped hemlock]

Hey NVIDIA GPU users. Are you using the "Game Ready Driver" or the "Studio Driver"? I just switched to the Studio Driver based on reading the use case from Nvidia. I'm not sure it made a difference. I'm thinking it may have made things worse. Anyone else using the Studio Driver? got any comments on this?

[ornate elk]

Make sure tiled vae is enabled

[ornate elk]

Hey, most users use the Game Ready driver

There is no big difference and I would also recommend using it

[warped hemlock]

Cool. Thanks for the reply.

[hardy canopy]

I have an In-Painting question. I'm trialing Forge, coming from A1111, and my in-painting isn't blending smoothly. There's a visible noise. I keep playing with settings, but can't get it to go away. Thoughts?

My settings:
Steps: 22, Sampler: DPM++ 2M Karras, CFG scale: 5.5

[ornate elk]

Are you using an inpaint model?

[hardy canopy]

I am not.

I've had success in A1111 with using regular models for in-painting. Was hoping the same would be true in Forge

[vocal burrow]

in theory no reasons for it to be different

try increasing steps. keep in mind that the "actual" number of steps in inpainting is defined by sampling steps*denoiser strength+1

[hardy canopy]

Do you guys know what's up with the resizing settings in in-painting? Should this be set to the size of the image that you're using as your base image? Or should this be set the size of the in-painted section that you're working on?

[vocal burrow]

That defines the size of the output.
I d say if you re only doing inpainting, set it to be the same as your input. And then upscale later on if you want.

Otherwise you re gonna waste time upscaling an image that might be badly inpainted.

And also you ll be inpainting on an image that is under reconstruction soooo the mask that you set initially around the face might end up being a mask for the face + whatever appeared during the upscaling.

[hardy canopy]

iiiinteresting. okay. okay. thank you kindly for the tips

[ornate elk]

Not its not the output (in his screenshot)

[ornate elk]

If inpaint area is set to only mask its the size of the inpaint area.
It has to be idealy a square resolution matching the trained size of the model. Sdxl = 1024x1024

Its like generating an image into an image

[hardy canopy]

Aaaa, okay. Okay. Thank you for the clarification

[ornate elk]

If inpaint area is set to "whole image" then it defines the output size like aryetis said

[karmic crown]

try to disable hardware acceleration in browser settings, that will reduce GPU usage and might help with performance. but it depends on browser and open tabs

[uncut jewel]

Heyo! Anyone had the issue with the SUPIR Conditionner "TypeError: 'NoneType' object is not callable" ??

I saw some open issues on Github on that matter, but no fixes until now

[vocal burrow]

that's a really generic error, can't help you without the full log and settings used to get that result

[uncut jewel]

It happens from on Workflow i found on ComfyUI workflow website, but aswell on the example workflow from the SUPIR folder

That's basicly this issue https://github.com/kijai/ComfyUI-SUPIR/issues/72

[vocal burrow]

try asking in #🧣｜comfy-ui and provide the full log there

[uncut jewel]

I'll do it, thanks!

[latent kelp]

Hi. Is there a link to solve the problem in photoshop with an error when using the inpaint sampler with karras?

[fair oxide]

is there a way to iterate through all LORAs/embeddings using scripts or something?
like you can with models?

[cosmic furnace]

@ornate elk Done some more testing with next?)

[ornate elk]

Nope

[cosmic furnace]

simple auto > next ? ;o

or next better only cuz noob friendly ?)

i see 🤔

[ornate elk]

next is less noob friendly than auto ^^

[cosmic furnace]

btw indeed tweaking fan curve & limiting fps (also lil bit lowering power) ended up 56 temp & 65c hot spot ;o

[ornate elk]

sounds good!

[cosmic furnace]

probably messed up smwhere with OC or dunno to have 91 hotspot =\

but i guess if max overclock its fine? hmgm

[ornate elk]

91? idk

[lunar orchid]

I'm debating a 3090 (used) vs 4070 Ti Super - I think that is a more 'appropriate' comparison since the other 4070s are only 12gb vram

no chance, guys, of a 7900 xt/x improving as an 'AI card?' 🙂

AMD seems interested in improving AI support - although, I guess that would limit one to using only Linux or? I have interest in using Linux but some ppl report issues using rocm - and that you can get crashes or problems? But, maybe that is fixed, eventually (dunno about soon)?

[ornate elk]

7900XTX is a good AI card on Linux with Rocm.
on Windows its now fast too thanks to Zluda release in february

[regal kestrel]

is this where i can ask one simple question about sdxl+lora little problem about speed I have or dont understand? or should I ask about it in sdxl channel? if you respond pls at me

[ornate elk]

Hey, Sure you can ask here, what's the issue?

[regal kestrel]

I got 2060 6gb and 16 ram and use a sdxl ckpt in forge. when I simply prompt I am ok with the speeds. But when I use loras it takes a lot of time. Am I missing something?

[ornate elk]

Only when Lora's are used?

[regal kestrel]

mmm

like 2-3+ min

[ornate elk]

For sdxl its better to increase the windows pagefile because 16gb RAM could be to less

[regal kestrel]

uses king haki

so I should fix it with a 8gb ram? or mess with the pagefile?

[ornate elk]

The pagefile increase is free so go for that first ^^

Make sure to also have 15gb or more free space on C drive

Then make sure the pagefile is only enabled for C, and disabled for any other drive

Set the size to 16000 min and 24000 max

Then reboot

[regal kestrel]

alright can I dm you the guide I see just to confirm the steps? dont want to spam here my issue

[fair oxide]

is there a way to iterate through all LORAs/embeddings using scripts or something?
like you can with models?

[vocal burrow]

what are you trying to do ?

[balmy lily]

if i wanna inpaint an image, and the part that gets inpainted i want to be based off another image, how do i do this, preferably in comfyui. so like instead of the masked portion getting generated based on the text prompt, its generation will also be based on another image i provide

[onyx solar]

Use IPAdapter+, encode the image, feed the mask and everything needed into IPA+ like you would an IMG2IMG, adjust weights & settings until you're happy.

[fair oxide]

If I have 100+ LORAs, I'd like to iterate through all of them. You can do this in a1111 using xyz script with checkpoint name option

[vocal burrow]

yeah I got that part. But what are you trying to do while iterating through them ?

trying to write a prompt replacing a world with your loras one by one ?

[fair oxide]

just test out different LORAs

I guess?

but when you have hundreds of LORAs, the filename itself is not necessarily how the LORA is called in A1111

"loraface1.safetensors" might be referred to as facelora1:1, for example

[vocal burrow]

Okay.... I'm gonna guess you want to have a prompt like
a cute xxxxx and replace do one output per lora, wherexxxxx is replace by a lora. Probably take a look at Prompt S/R https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features#prompt-sr.
And get a list of your lora files through by running this in your Lora folder

for /R %f in ("*.safetensors") do @echo %~nf | find /v "ECHO is"

Armed with a list of your lora filename and Prompt S/R you should be able to something like I described before

(just make sure that your lora are set to activate using their filename instead of aliases in a1111 settings)

[radiant arch]

So I just got my PC set up, a rig with a 4080 Super in it. I assumed that I would still be getting better performance with the Forge webUI. Was I mistaken?

[dapper pine]

how do i fix this?

[onyx solar]

https://phoenixnap.com/kb/install-pip-windows

[fierce hollow]

Hello everyone, I know what may be at the wrong address, but here I believe there are many neural network specialists, probably someone can. I want to start making cartoons, please tell me if there is a neural network (it is very, very desirable that it be locally in the SD VAC) and could perform the task of character animation

I really hope that someone will be able to help and show something, thank you in advance

[obsidian parcel]

How do I get started?

[wet flame]

hey is forge ui better that automatic 1111? i got 4 gb of vram lol

been us A111

is forge better for low end gpus?

how to enable these

[oak sun]

hi guys ..i have just one question

is civitai helper slow or something is wrong with my auto1111

its taking long time to get info and png thumb for models ..is that on civitai end or from my end?

[ornate elk]

I think the speed is based on the Civitai site itself

[formal trout]

for curiosity, what is this?

[ornate elk]

A setting that got disabled in the new webui version.
In the settings you find it under cump

With an short explanation

[formal trout]

oh yes it shows with old images generated in previous versions

ty

[fierce hollow]

Hello everyone, I know what may be at the wrong address, but here I believe there are many neural network specialists, probably someone can. I want to start making cartoons, please tell me if there is a neural network (it is very, very desirable that it be locally in the SD VAC) and could perform the task of character animation

[formal trout]

here I found it

[formal trout]

good thing is under cump and not under something short

[grim iron]

still get the same error

[ornate elk]

Can you show your img2img settings?

Also make sure scripts are disabled at the bottom

[ornate elk]

@grim iron Try set the masked content to fill.
Then set the resolution to a square.
If your using an sdxl model to 1024x1024

[grim iron]

the original image is a rectangle so it will be squised in if I use 1024x1024 any way to fix this?

[ornate elk]

It won't be squeezed as you have Inpaint area set to "only masked"

[grim iron]

ohh ok

[grim iron]

still getting the same error

[ornate elk]

Can you show the cmd log?

[ornate elk]

Any extension enabled ?

Like controlnet or AnimateDiff ?

[grim iron]

installed by not enabled

[ornate elk]

Okay, and your using sdxl or 1.5 ?
Any loras used?

[grim iron]

sdxl no loras

[ornate elk]

Try restart the webui.
And then only select inpaint area masked only.
And 1024x1024

[grim iron]

nothing changed

[fierce hollow]

https://imgur.com/a/b5a3qQo
What are the effects of these items and how best to set them in order to add the maximum possible transfer of a face from a photo to the generation?

[brittle lantern]

£600 vs £720+ for me if I compare 4070 ti super

[fierce hollow]

Help. What's wrong with your eyes? How can this problem be solved?
Model: Juggernaut XL Speed skating as in any other CFG 2, Steeps 6 model
Promt:A girl as tall as wormwood, in a sweater and jeans, looks at the viewer, realism, detailed skin, ultra-realistic, natural lighting
Negativ: bad eyes, computer graphics, airbrushing, plastic, deformation, watermark, strabismus

Model: DPM++ SDE

[ornate elk]

Whats wrong?

Eyes can be improved with Adetailer or Codeformer Face restore

Or by upscaling

[fierce hollow]

Look at the eyes, they are different, I often meet people here with problems with their eyes, especially if a person is not standing straight, but for example sideways

[fierce hollow]

Is there a problem with evil eyes because of the small resolution ? (832x1216) ? And if I turn on Hires. fix - the problem will go away, right?

[fierce hollow]

@ornate elk Sir... We have a problem... -_-

When I started working in the neural network, I thought that I would get pleasure from it, not suffering...

[drifting pendant]

Hello, does anyone utilize all AMD hardware here ?

[ornate elk]

Yes! Me and a lot others do

[ornate elk]

Set the denois to 0.5 and hires steps to 10

Then it will work

[fierce hollow]

I'll try it later, thanks. You always help, if I had your photo, I would frame it and pray! You're a saint! Thank you very much

[lunar orchid]

120 more for the 4070 Ti S than vanilla 4070?

[drifting pendant]

Nice, I have a lot of troubles trying to run SD on my 3900x / 32Go RAM / 6800XT rig, I will try one last time to run Ubunto 22.04.2 LTS in one hour, last time I've tried, everything was good till I tried to render something, everything crashed, starting by the firefox instance 😂
I was about to give up my AMD graphics for enevrything else NVIDIA 😅
Il tell you if I succeeded 😉

[brittle lantern]

Tbh AMD tends to be completely fine for the most part on Linux with ROCm6 windows tends to be the issue with SD and AMD

Nvidia as much as I don't wanna go I have to I prolly do just cuz I use windows but the prices is so scammy from nvidia

[drifting pendant]

I'm brand new into this kind of stuff, and I have discover that my brain is not braining with this stuff 😅

[drifting pendant]

Sadly true 😭

[brittle lantern]

Dw I was like that also

• dual booting linux causes me more issues than just running Linux but I play games with anti cheats that don't on linux

[drifting pendant]

Anyways if it doesn't work, ill try Windows again 🤔

[drifting pendant]

Yeah, I have a completely different SSD for SD 😂

Don't want to dualboot Windows, because, Microsoft y'know

[brittle lantern]

I don't wanna dual boot cuz it causes issues

As a comp player I am forced on windows

[drifting pendant]

Yeah, classic Windows stuff 😂

BTW there anything "renderable" with a GTX1650 4Gigs ?

[brittle lantern]

Ita 4gb of vram I don't think so

Sd 1.5 will prolly be a issue with 4gb of vram

[drifting pendant]

Ahah thx for the info ! 😉

[ornate elk]

Do you want to use Linux or do you want to use stable diffusion on windows?
Cause on windows its now easier to get it running and a good performance

On Linux you would get the best performance but its not as easy to get it working and keeping it working

[ornate elk]

Yes that card will work too with SD

[cosmic furnace]

Did someone run into problem like browsers use too much CPU ? Espec Chrome? Like 4 to 6 tabs can get cpu usage up to 75-85% espec if big & with image

[ornate elk]

Make sure you don't leave Civitai open as tab xD, some websites can cause high ram and CPU usage

[cosmic furnace]

Either recent update of browser or whuht? At some point firefox do this too but more rarely & you need to try hard. Old cpu didnt had this issues at all =\ . & started noticing it recently cuz cpu cooler starts being loud when surfing -_-

Ive made an negative PBO curve (-25) and -0.1v overal but dont think its the case ? Also all my cpu tweaking doesnt seem to affect voltages of CPU if looking at overall usage . Maybe core voltage is smthing different?

Its just weird that alot weaker prev CPU didnt had this issue with huge CPU usage. And not only civit AI . Its just seems like any tabs do this

@ornate elk What browser dyu use btw?) And what dyu think r best

[ornate elk]

I use Firefox with ublock origin
Works perfect for me

[cosmic furnace]

smone advices to disable inheritance in folder of Chrome Software Reporter Tool . So it cant acces it xD

or clear cache -_- . I hope its not 5700x issue? doesnt look like all browsers r like this 😒

Y i guess have to switch over to firefox & opera =\

[ornate elk]

Maybe you disabled Hardware acceleration in chrome by accident?

[cosmic furnace]

got used to chrome but ...

nah its on... and ye tried to disable with it disabled cpu usage goes even higher xD

[ornate elk]

Wouldn't recommend Opera

[cosmic furnace]

Opera gx then ? )

u wanna say edge better than opera? 🤨

[ornate elk]

No any opera is bad cD

Especially GX with its bloated addons

[cosmic furnace]

what ar alternatives? 🤔

[ornate elk]

Brave, Vivaldi, Firefox

[cosmic furnace]

those ar best browsers? brave & vivaldi first time i hear ;o

[ornate elk]

Brave is good for privacy and as fast as chrome/opera
Vivaldi is made from the Opera creator as 2016 Opera was bought of by a Chinese consortium

[cosmic furnace]

do they "eat" chrome extensions?)

[ornate elk]

All work with chrome extensions, only Firefox has its own

[hushed knot]

I love Firefox w/ UBlock Origin <3

[cosmic furnace]

i need to have few browsers +) . Or even 3 of them xD

currently have 4 if i count edge >_>

hmgm

[ornate elk]

My focus is mainly privacy so I even uninstalled Edge from windows

[drifting pendant]

Oh thanks, I will run some simple renders on it then, thx ! 😊

[ornate elk]

Checkout my install guides. To get the best performance settings.
The guides are in the pinned messages of this channel

[fierce hollow]

It helped, thanks!!!!!!!!!!!!

[ornate elk]

Perfect, no problem 🙂

[cosmic furnace]

Btw is zluda version of auto have any limited functions? Well something it cant do ?

as of now)

[drifting pendant]

Honestly I'm looking for the most performing option and I don't like Windows that much so linux it is eheh

[drifting pendant]

Oh thanks !

[ornate elk]

99% of things work on it. There are some extensions that won't work as they rely on Cuda DNN. The Zluda DNN support is currently work in progress.

[drifting pendant]

BTW excuse my French, I'm actually one

[lunar orchid]

Who rated those? LOL!

[cosmic furnace]

Chromium browser good ?)

Its better to download full model then pruned y? hmgm

[ornate elk]

Some are good some are bad

[ornate elk]

Nope, pruned are smaller and load faster, also they require less vram

Working better with upscaling

[brittle lantern]

I'm glad Edge is on that cuz it gets to much hate

[brittle lantern]

I mean Edge is correct in a way it's the most optimised by per windows build tbh

[fierce hollow]

By the way, it doesn't always help, there are artifacts (I grow new body parts :)), I had to play a little with the values , denois to 0.4, steps 15) It seems to have worked

[cosmic furnace]

Some nice graphics AAA games cant load cpu like chrome does 😵‍💫 . Sadge but its time to say goodbye to chrome i guess =\

[drifting pendant]

AMD folks, it is require to disable the SVM mode when you are using ROCm 5.x (linux) on your system ? Just to be sure 🤓

[cosmic furnace]

Btw someone did testing about on what browsers SD works best? Or doesnt matter ?)

[ornate elk]

yea you have to to play around with the settings, also its differs for the upscalers

[ornate elk]

doesnt matter

[gloomy pivot]

Hey Everyone!
I've got a quick question, I've seen many tools where you can upload interior or your photo and they use some kind of AI to do stuff like changing the background, removing acne on someone's face, changing furniture etc.

Does anyone have an idea how these are done? We want to build something similar for our internal use, but I honestly have no idea as to where to start.

Thanks!

[teal arch]

hey i want to istall comfy what python version is optimal?

[ornate elk]

hey, python 3.10.11 64bit

[teal arch]

thanks

[lament fog]

I run it, but I have no idea what that means. I'm just using a pre built docker image

[merry cairn]

can someone tell me what node to use in comfyui to work with 1.5 models ? KSampler doesn't seem to work

[open fox]

the default workflow should work fine

and it does use the basic ksampler node

[ornate gorge]

Why would this degrade? I was usingsame settings as before and before my animating with auto1111 + deforum plugin were working, I have even tried to change resolutions to match to xl etc and it's still deforming and shitting like that after 10-20 frames, any help apprecieated

or is it sampler issue?

[ornate gorge]

keeps happening with different samplers too, idk, few weeks ago i was able to animate no issue with same setup

[thorny iron]

What does this mean

[thorny iron]

I cannot replicate my past works either. It's all slightly different.

It's as if some priority or sequence has been tweaked and affected everything

[dapper pine]

still didnt work, or i installed it wrong

[graceful girder]

hello ! I would like to know how to create my own lora, does stable diffusion have a function for it and are there any prerequisites to have or other? (sorry if it's the wrong one Chanelle I didn't really know where to ask for help ^^')

[quaint sluice]

Hello, just a question here, for AMD Radeon do we have to use ONNX or we can use Stable diffusion ? My 7600XT with 16GB is not working properly with Stable Diffusion. Thank you

[ornate elk]

Hey, on windows the best for AMD is to use the Zluda webui version.
You'll find a guide in the pinned messages of this channel or here:
#🤝｜tech-support message

[light hill]

Hey everyone,

I need some quick help with the Stability.ai inpainting API endpoint. I'm sending an image and mask according to the API documentation, but I'm noticing that some areas not included in the mask are still being inpainted by the model. Has anyone else encountered this issue, or is it just happening on my end?

[hardy rapids]

when doing batch generation how do i know which image has what seed? if i set to random do they all have random seed or only the first one is a random seed and the following ones increase seed by 1?

[ornate elk]

In the meta data the seed is shown.
Normal the first is random, then it increase by 1

[sudden hawk]

hello, i had this problem 3days ago with a check point : to have images fully grey, I didnt success to fix it, and now I have these grey images on every checkpoint I use.
I have this in my webui
*** Error running install.py for extension C:\A1111\stable-diffusion-webui\extensions\sd-webui-roop.
*** Command: "C:\A1111\stable-diffusion-webui\venv\Scripts\python.exe" "C:\A1111\stable-diffusion-webui\extensions\sd-webui-roop\install.py"
*** Error code: 1
I have rtx 3070 and use SD 1.8
could someone help please ?

[ornate elk]

Delete the roop extension, then delete the venv folder

[sudden hawk]

ok, i have desactivate the roop extension on SD interface but I don't understand about venv folder to delete > you don't mean the one in A1111 ?

[ornate elk]

Don't disable the extension, you need to delete it from the extension folder. Because roop is outdated and broken.
Then delete the venv folder inside the stable-diffusion-webui folder

Then relaunch the webui-user.bat

[alpine harbor]

hi all. Does anyone here have Stable Diffusion running on Linux "Ubuntu", I have run into a runtime error that i cannot seem to resolve.

[sudden hawk]

ok, so it's done, i've deleted sd-webui-roop and venv folder and now I have this error msg in my webui ImportError:
cannot import name 'Undefined' from 'pydantic.fields' (C:\A1111\stable-diffusion-webui\venv\lib\site-packages\pydantic\fields.py)

[ornate elk]

Ohh that error appears when using a not updated controlnet version.
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

[sudden hawk]

I did, i have also relaunch webui, and I still have grey images, and now i have this in my webui
*** Error loading script: main.py
Traceback (most recent call last):
File "C:\A1111\stable-diffusion-webui\modules\scripts.py", line 527, in load_scripts
script_module = script_loading.load_module(scriptfile.path)
File "C:\A1111\stable-diffusion-webui\modules\script_loading.py", line 10, in load_module
module_spec.loader.exec_module(module)
File "<frozen importlib._bootstrap_external>", line 883, in exec_module
File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed
File "C:\A1111\stable-diffusion-webui\extensions\openpose-editor\scripts\main.py", line 14, in <module>
from basicsr.utils.download_util import load_file_from_url
ModuleNotFoundError: No module named 'basicsr'

[ornate elk]

Openpose editor is also outdatet you can delete it from the extensions folder and relaunch

Then in the extensions tab click on Check for Updates and let them update

Then relaunch

And then show me your settings when you get the Gray image

Like resolution, model, steps etc

[sudden hawk]

I did all you say and still have grey images

[jolly wave]

create a nice Cut to reenactment of ancient Egyptians working on the construction of the pyramids

[ornate elk]

And you have --xformers --medvram-sdxl --no-half-vae in your webui-user.bat?

[sudden hawk]

yes

[ornate elk]

Okay then close the webui.
And delete the ui-config.json
Then relaunch

[sudden hawk]

Thank you so much it works now !

thanks for your time again

[ornate elk]

No problem 🙂

[fierce hollow]

Here are the permissions for SDXL
aspect ratio: 4:3 - 1152x896
aspect ratio: 2:3 - 832x1216
aspect ratio: 9:16 - 768x1344
aspect ratio: 21:9 - 1536x640
aspect ratio: 1:1 - 1024x1024
aspect ratio: 3:4 - 896x1152
aspect ratio: 16:9 - 1344x768
aspect ratio: 5:4 - 1152x896
aspect ratio: 3:2 - 1216x832
aspect ratio: 4:5 - 896x1152
aspect ratio: 9:21 - 640x1536

Is there a similar list for 1.5?

[mortal compass]

https://imgbb.com/qkHKcnp

Anyone here familiar with Interrugate CLIP/Deepbooru? I'm wondering if its possible to batch interrugate images. Got loads of images for a lora and doing it 1 by 1 will take days 🤣

[ornate elk]

Kohya and onetrainer have an automatic clip deepbooru tool that will run trough your image folder

[cosmic furnace]

@ornate elk Issue with sd next 😒

[ornate elk]

Wrong vae

[cosmic furnace]

8 or 10 hours later no other answers than y wrong vae but...

and try restart sd next

[mortal compass]

Thank you, google gave me nothing before, I'll check them out!

[mortal compass]

Put it on automatic

[cosmic furnace]

got asked what model is . And answered that this one . https://civitai.com/models/317902?modelVersionId=469115

And the most funny part is , i tried running it on zluda auto 😄

and it worked. no issues.... right away ...

[ornate elk]

Ah well, then its an sdnext setting issue

But why keep using sdnext when you have auto zluda xD

[cosmic furnace]

while on next didnt =\ ... gotta agree i guess that maybe it looks that sd next has more inbuilt functions and bigger UI. It is frustrating to use for a newbie... U gottta ask what do what there... Also maybe im wrong but it looks like usual zluda auto runs faster

Because even when launching usual zluda auto command line says that zluda works best with sd next ? 😄 . And use it there better ? -_____-

[ornate elk]

Yes sdnext has to many stuff integrated and works different in the back. Also the worst Ui xD

[ornate elk]

That recommendation is only because the auto zluda dev helped with sdnext too

[cosmic furnace]

Btw in built styles doesnt seem always to work jeesh. And this latest pony model is so good and smart o.0 . No wonder all ppl praise it in comments ;o

[ornate elk]

But he still works on auto with zluda, so I would recommend using that over sdnext

[cosmic furnace]

rly weird why i cud run 1.5 prev model on next and not this one. Btw nex doesnt have 3pm karras for some reason. I tried restarting both with auto vae and non vae there, maybe its hard for him to deal with inbuilt vae ? o.0

Does you pinned message

got all settings need to set up for it ?

[ornate elk]

Yes sdnext uses a diffusers Backend. Thats why it misses some samplers and maybe doesn't work good with pony models

[cosmic furnace]

Zluda next guide has some experimental & optional settings . That might run things better-faster

[ornate elk]

My Zluda guide is complete yes

[cosmic furnace]

r u sure? xD

maybe theres something i can tweak to run it better-smoother ?

[ornate elk]

It already runs smooth lol

I got no issues with it

[cosmic furnace]

although gotta tell . This sdxl & non turbo model feel like runs quite fast hmgm

1560-1280 generation in about 30 sec is good y ?)

[ornate elk]

Definitly yes

[cosmic furnace]

y 1280-720 takes about 7 sec ;o

[ornate elk]

Because of the lower resolution

[cosmic furnace]

and its non turbo-hyper-lightning version. Why so many of them -_-

And how do i know if its sdxl or 1.5 when author doesnt specify it ? By size? o.0

[ornate elk]

Check the info box

Above

There should be stated 1.5 or sdxl

[cosmic furnace]

This pony is so good as if they got alot of inbuilt loras and poses. ) . One 1.5 lora was always doing same pov and pose . But pony boom ;O

its just says "other" 😄

[ornate elk]

Oh okay

[fallen quest]

SDXL is not even a year old and the model was verified a year ago

[ornate elk]

But thats an 1.5 based model.
Sdxl models are always larger than 6gb

[cosmic furnace]

Ive heard an opinion that you can overtrain model and it ends up more bad than before? Like newer versions can be worse sometimes? 🤔

[ornate elk]

Yep, that can happen to loras too

But loras luckily can be adjusted in strenght

[cosmic furnace]

So this inbuilt vae pony problem might be just unsolvable for next huh ? 😒

[ornate elk]

Try use it with an sdxl VAE and it could work

[cosmic furnace]

Can i link models from next to auto zluda? And will it be slower that way or some issues ?

I d like to have both for now and not. Moving models from 1 to other)

Also is there any point in keeping usual directml now ?

Zluda seems to run fine & fast hmgm ...

[ornate elk]

Yes you can link them and it won't slow down.
Edit the webui-user.bat and there add:
--ckpt-dir "Path/to the models" --lora-dir "path/ to / loras.."
--embeddings-dir ...
--controlnet-dir

[ornate elk]

Nope, but I still have it installed. But I deleted sdnext after my tests a while ago

[cosmic furnace]

:DDD

Didnt deleted kinda "useless" directml. But to prove that Next is even more useless just deleted it xDD

theese goes to COMMANDLINE_ARGS y ?

It will use both its own dir and specified or only specified hm ?

[ornate elk]

It will use both

[cosmic furnace]

Btw maybe i dont know where to look but Next sometimes doesnt have added embeddings & loras in UI . So ull have to use them with writing -__- . Dunno if its zluda or usual next also 😒 . But sadge

so far i guess indeed in overall score next version kinda loose to usual auto 🤔

[ornate elk]

That has nothing to do with zluda

[cosmic furnace]

Was even looking discord history to make sure its still an issue. And turned out it is. Ppl just assumed zluda at fault cause ofc i was searching zluda channel)

[ornate elk]

@cosmic furnace Zluda is like Directml or onnx or cuda (in fact it emulates cuda)
So its just a Backend of the webui that tells your GPU how to process with the data.
Everything in usage, overlay specific is webui based.

[cosmic furnace]

Choosing pony's sdxl vae didnt do the trick 😦

[drifting pendant]

Hello, I have finally have SD running on my PC, on Windows, not Linux as I excepted, but it is running fine, I have a 6800XT and my GPU memory is running out of hand, except I don't use models over 7Gb, what is causing the issue ? :/
PS : I have ROCm installed too

[cosmic furnace]

model size doesnt equal vram usage)

at some point memory can just run out. Ive heard even nvidia have those issues hmgm.

[drifting pendant]

Oh, ok Thx mate, is a virtualization option to run Virtual machines on your system 😉

[cosmic furnace]

What resolution you generate? ;o

[drifting pendant]

--mid-vram then ?
512x512, sometimes upscaled and 1024x1024

Nothing crazy

[cosmic furnace]

weird. medvram y helps). From what i understood even 7900(any) is recommended to run with medvram 🤔

[ornate elk]

Have you used my Zluda guide?

[drifting pendant]

Nope

Maybe I should ^^

[ornate elk]

Then you need to do that for the best performance + less vram usage and faster image generation

[drifting pendant]

Let's go then

[ornate elk]

Here it is:
#🤝｜tech-support message

[cosmic furnace]

Oh it was just directml . >_> and i thought on zluda ;o

[drifting pendant]

Since I have a version of stable diff already installed, I just have to delete the actual buggy one ?

[ornate elk]

Yep, delete it. Its the direcml one which is very vram limited and slow

Don't get confused as my guide also needs the files for preparation. So just follow it step by step

But better start clean

[cosmic furnace]

Btw interrogating image on next didnt worked... gave it 300(or even more) sec to try =\

Is there a good stable way to get an more less good text promt by image now? Well when not specified in info

[drifting pendant]

Alright I'll try it, thanks again ! 🙂

[ornate elk]

NP, let me know if anything isnt clear 🙂

[ornate elk]

Not really

There is a interrogate extension for Auto with different clip models

[oak sun]

Hi

How do I run kohyass which requires python 6.11 along with auto1111 which requires 6.10 together on a pc

[ornate elk]

Kohya normal should work with 3.10.11

And auto1111 too

[cosmic furnace]

sometimes randomly just get this error and cant generate anymore until restart =\

I dont get this using next btw 🤔

Hm 🤨 . Started to get this error often weird.

[limpid nest]

Hi, i need help with that, i generated a prompt with a protogen models and i obtained this, can you help please ?

[cosmic furnace]

again error right before ending. Mb shudnt use 1.5 loras with Pony types -_-

[onyx rivet]

yo guys how do i install it on linux (amd)

[oak sun]

So this outdated

Also I'm still on 1.7.0 release so I don't have the latest torch .. I don't know how that will work

[onyx rivet]

what about the native(linux) tutorial

[oak sun]

Wdym ..I'm on windows

[onyx rivet]

oh

im tryna install on linux

[ornate elk]

You should update your webui and torch too then

If you want

[onyx rivet]

RuntimeError: Couldn't clone Stable Diffusion.
Command: "git" clone --config core.filemode=false "https://github.com/Stability-AI/stablediffusion.git" "/home/meezopc/stable-diffusion-webui/repositories/stable-diffusion-stability-ai"
Error code: 128

uhuh

[ornate elk]

You may need to update your zluda files

[oak sun]

There is a reason I don't want to update it

In staying on 1.7.0 as I don't want to update my nvidia driver anymore

So my question is will 1.7 work with python 3.11 or should I find older commit of kohya ?

[ornate elk]

Kohya doesn't need 3.11

It works fine with 3.10.11

Thats the best version you can install for ai stuff. It works with everything

[oak sun]

Yeah I downloaded it right now ..I'm gonna see if my auto1111 commit works on it ..thanks

What is the preferred model for tagging rn?

[cosmic furnace]

@ornate elk Csssiiiiiiioooooooo ;o

[ornate elk]

Depends, for anime its still boorutag

[cosmic furnace]

https://youtu.be/l-uAjD2s9jE 🙂

Oh zluda files update? The one in zluda folder?

[oak sun]

Is lora trainer colabs or did Google shut it down??

[cosmic furnace]

Pretty sure i did recently 🤔

[ornate elk]

Yes

[oak sun]

OK..I'm not interested in anime really ..I haven't touched SD in a year almost but I used to use blip mostly

I heard there was a blip 2?

Nvm ..I'll just try them all and see what works

[oak sun]

Do they still work ?

[cosmic furnace]

No thats not it still happening =\ .

Very first or second try now ends up on error on 96% completion

almost nothing in command line o.0

Mb because moved model? Kinda doubt it)

wait what o.o

why it uses zluda here ? it does or it lies?

[junior heron]

type cls

[cosmic furnace]

where?)

when running sd commands not working for some reason hm

but i can tell it does because it doesnt let me move em while i run it -_-

Yes i updated zluda files in that specific folder and no more errors... Shud had specify it. Its weird because via guide i was installing zluda MY folder

the one i specified on C drive

And all of a sudden recent install of it and it has a folder of its own inside and he uses it instead o.0

😵‍💫

[ornate elk]

yea the dev changed the installer

it now downloads its own zluda files but on 3.5,
3.7.1 are the latest

[cosmic furnace]

got this huge error again when upscaling in img2img 🤨

[ornate elk]

show you upscaling settings

[cosmic furnace]

well just resizing xD

[ornate elk]

and what version is your webui?

[cosmic furnace]

webui version? o.o

[ornate elk]

at the bottom

[cosmic furnace]

this?

[ornate elk]

yep

your one update behind it seems, but nothing major changed

[cosmic furnace]

😑

smthing can be done? 🫠

[ornate elk]

what was the source resolution?

[cosmic furnace]

i was installing via yur guide... recently and ye i remove -update from com args)

1024

[ornate elk]

--update-check isnt for auto update

for that you ned git pull

[cosmic furnace]

;o

where do i put this git pull ?)

adding to com args not working

ah i see cmd

[ornate elk]

but only add it if you want to automaticly update

you can also remove it after running it once to just update once

[cosmic furnace]

Made few imgs after updating and error again 🫠

is there some kind of log txt file with errors? 🤨

[ornate elk]

nope only the cmd

but when do you get the error?

only when upscaling?

img2img or txt2img?

[cosmic furnace]

y lately was trying jsut upscale from 1024-768 to 1280-1024 img2img with resize

to get more details & higher res 🤔

indeed it seems just txt2img i can generate no prob o.0

lower-higher res

wtf ... or maybe its 1.5 loras i use in prev promt?

img2img is still problematic? My friend with weak nvidia told he also gets few upscales and then error o.0

[ornate elk]

you shouldnt upscale with "resize to", you should upscale with the sd upscale script at the bottom

[cosmic furnace]

whaaat o__O

?

so many options? o.0

ah indeed sd-upscale at the bottom)

erm this looks more complicated than without 😦

isnt there stable just resize option? xD

[cosmic furnace]

indeed with upscale script seems no error ^_^

But i get weird overlap or image divided into sections o.0

[ornate elk]

show me your img2img settings

with sd upscale script

[ornate elk]

you need to set the denois below 0.3
then set a square resolution

[cosmic furnace]

feels like managed to get image that looks fine.

so far more less like this. but maybe because alt tabbed too much got error after 3 times =\

hm like this 5% of the image from left overlap& doesnt match o.0

i cant or shud change size of image to higher ?

or only scale

and what if i wanna see denoise variations with slightly higher res ? 🤔

and error o.0 after 1 time

hhmmm ?

or change browser they say lol -_-