🤝｜tech-support

[ornate elk]

if you only have the Intel UHD gpu,
then you need to add: --use-cpu all --no-half --opt-sub-quad-attention
then save.
then delete the venv folder
then launch the webui-user.bat

yes

seperated by space

[ornate elk]

have you deletet the venv folder before?

[ornate elk]

Okay can you show the error log ?

[ornate elk]

Okay, yea the install part can be Laggy on Laptops

Mostly you have only 8gb of RAM

[worldly rain]

hey where can I download the models for ip adapter sdxl?

I refreshed it I dont have any installed

[ornate elk]

https://huggingface.co/h94/IP-Adapter/tree/main/sdxl_models

[worldly rain]

thanks a lot

I download bin files only right

[ornate elk]

no, download the .safetensors

[worldly rain]

okay ill download all of them

[ornate elk]

yea it has 4 versions/models

[worldly rain]

where do I put them tho

[ornate elk]

in the models/controlnet folder

[worldly rain]

much appreciated

[brisk heath]

what's Invalid HTTP request received error and how do i fix it

[vocal burrow]

try updating your extensions

what's your gpu ? and what are your command line args ?

[brisk heath]

i just did, and when i checked again it says everything's at the latest version

[vocal burrow]

then I don't think there's anything to do. It's probably due to some extensions doing some stuff wrong.

[ornate elk]

hey, whats your GPU?

[serene robin]

6700 xt 12 gb, and what's a command line arg

[serene robin]

amd 6700 xt, i remember you helped me before to set it up

[vocal burrow]

what's in your webui-user.bat file ?

[ornate elk]

since the latest version you need to add --use-directml to the webui-user.bat

[serene robin]

git pull
@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --medvram --no-half-vae --opt-sub-quad-attention --opt-split-attention-v1

call webui.bat

[vocal burrow]

add --use-directml to the set COMMANDLINE_ARGS= --medvram --no-half-vae --opt-sub-quad-attention --opt-split-attention-v1 line

[brisk heath]

turns out it's not the extension, since i turned them all off but it's still happening

maybe it has something to do with me downloading pytorch on the side for upscaler

[serene robin]

it worked

thank you @vocal burrow & @ornate elk

[glossy prairie]

Will do, thanks!

[worldly rain]

guys why is my openpose is not working? its not effecting the generated image at all, in the preview it shows something irrelevant too

[bronze leaf]

Working again!

[ornate elk]

the preview seems correct, make sure the controlnet extension is updated

[worldly rain]

it is updated i checked it

[ornate elk]

try disable ip-adapter and then try again

[worldly rain]

that didnt help

but when I tried to install dw_openpose today it gave me an error

i dont see dw_openpose in the installed extension section but I see it in the preprocessor

the error goes like this

LOL

when i try to install it again it says

GitCommandError: Cmd('git') failed due to: exit code(128) cmdline: git clone -v --filter=blob:none -- https://github.com/IDEA-Research/DWPose.git C:\Users\Enes.EMRE\Desktop\stable diffusion\stable-diffusion-webui\tmp\DWPose stderr: 'fatal: destination path 'C:\Users\Enes.EMRE\Desktop\stable diffusion\stable-diffusion-webui\tmp\DWPose' already exists and is not an empty directory. '

maybe it has something to do with this eror

[wet surge]

Hi, i downloaded a model from civitai, and placed it in the models/stable diffusion folder. What can I do to make it work on windows 11? btw this is the error code: 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]

The error is that you installed DW openpose at all.
Its not needed as its included by default in controlnet

[ornate elk]

Check if the file is larger than 2gb
If not then its not a model and in the wrong folder

[worldly rain]

okay i'll delete it then, maybe my normal openpose can work that way as well

[ornate elk]

Yes delete the Extension.
Ohhh and I see the error "no space left on device"
So yea... XD

[worldly rain]

lmao

i got like 6.8gb left in my c drive

XD

thanks a lot

[wet surge]

Thanks! But then can you tell me what is this model safetensor?

[ornate elk]

xD make sure to have 15gb or more
So windows and its Programms can work normaly

[worldly rain]

🤝

[ornate elk]

Thats a lora.
Loras are small files trained on a specific subject/style/character
These are used together with an actual model to get the wanted result.

Loras go into models/lora folder

[sonic ferry]

I'm using this workflow I found on Civitai, basically the same setup I was using except the CR Apply LoRA stack between the Loader and the KSampler. My question is, how do I add LoRA that needs a different ClipSkip without affecting the rest? I figure leave my LoRA connected to the Efficiency Loader's left, without changing ClipSkip there, but how do I add another LoRA stack that does need ClipSkip 2?

[sonic ferry]

Well, their Clip from the Apply LoRA Stack isn't going out to anything, so I figure this should work for some reason?

[grand fable]

Hi, do you have some experiences with running ComfyUI on AWS (Amazon cloud computers)? I have only M1 macbook so this could be good alternative for learning SD..

[sonic ferry]

Not I, I'm running ComfyUI on a calculator.

[white gull]

HI, I am fairly new to AI in general but very very new with Stabile Diffusion. I was using a site that allowed me to create a custom model using Dreambooth and it worked great on that site, https://getimg.ai/. I was able to download the model and tried to use it on a local machine. But I have not been able to reproduce the results that the website was able to generate. Not even close. It is as if I am not using the correct model at all. I feel like I just don't have things where they need to be. Or more likely, I am just not understanding what I need to do to use it properly. Can anyone shed some light on this for me?

[karmic crown]

how big is your downloaded model?
what was the base model for training?

[white gull]

The model was over 7gb, I got the safetensor and the chkpt and a bunch of others I dont know anything about.

[white gull]

The base model was ICBINP was the base I chose.

[karmic crown]

ok, that its SD 1.5 model

what setting are you using on your local machine?

dimensions, sampler, steps, cfg...

[white gull]

I have tried so many I cant remember them all. Can we start with where the files I downloaded need to go? Should I be selecting it in the checkpoint selector or in the lora area? are they entriely different?

[karmic crown]

that's why i asked about size

[white gull]

I dont even understand exactly what I have

[karmic crown]

2GB and more, its a model
and goes in models/Stable-diffusion folder

[white gull]

I keep it at 512x512, I dont have a beast of a GPU

and a model is the checkpoint selection at the top left in webui?

[karmic crown]

yes

[white gull]

OK, So I choose that and if I have it all set up correctly, anything I generate should reflect the model I trained?

[karmic crown]

from ICBINP page

Recommended settings: DPM++ 3M SDE Karras or DPM++ 2M Karras, 20-30 steps, 2.5-5 CFG (or use Dynamic Thresholding), happiest at 640x960 with a hires fix on top, but if you are happy to hunt through seeds to avoid head duplicates, 768x1152 works quite well also.

so try this

512x512
DPM++ 2M Karras
cfg 5
steps 20

do not use any loras, hires fix or anyting else

just test the plain model first

[white gull]

OK, I need to rearrange some files. Do I just keep all of that in the folder structure I downloaded or can I keep just the 1 file I need and dump the rest?

[karmic crown]

show me the list of files you downloaded

[white gull]

the pngs were added by me and were not part of the download

[karmic crown]

just put "Lara V2.safetensors" in Stable-diffusion folder

[white gull]

everything else is just bagage?

[karmic crown]

keep it, but do not put it there

[white gull]

ok

I dont need a corresponding json file?

[karmic crown]

you can put jsons and preview.png

but Lara V2.json, not model_index.json

[hoary trout]

so
i've been having a rather persistent and vexing bugbear of a problem
sometimes when trying to generate an image my pc will just flat-out freeze
which then requires me to restart my whole pc
this has happened twice in the past day
upside is, i'm pretty sure i know what's causing it
downside is that i think it's not something i can easily solve

[karmic crown]

@white gull select your model

use those settings

512x512
DPM++ 2M Karras
cfg 5
steps 20

and simple prompt like
1girl

[white gull]

And were back:) OK, Ill give that a shot. Thanks for your help.

[ornate elk]

hey, that issue appears if you have only 8gb of RAM in your PC. Or if you have 16 and your C drive is full

What you can do is to increase your windows pagefile and make 15gb free space on C

[hoary trout]

i have 16, but my c drive is actually only about a third full

i have about 600 gigs free

[ornate elk]

okay and whats your GPU?

[hoary trout]

also i feel it pertinent to say that this only happens sometimes

[white gull]

So that is actually getting things a little closer. I want to keep the prompt simple to try to fine tune the model...I think. What adjustments should I try to pull more detail out of the model? I assume it is either sampling steps or cfg? I understand cfg is how strongly it sticks to the prompt itself. I dont really understand sampling steps yet.

[ornate elk]

okay, and whats inside your webui-user.bat?

[hoary trout]

the first time this happened was before it was in this state, but here

it only had the first three iirc

[ornate elk]

you can remove --medvram and add --use-cpu all --opt-sub-quad-attention

[hoary trout]

ok
would this make gens take less time?
as it sits, gens take me upwards of an hour

[ornate elk]

yea that should work faster, if not then remove --precision full and add --upcast-sampling

[hoary trout]

will this also help with the russian roulette freezing issue?

[ornate elk]

yes hopefully ^^

[hoary trout]

welp, here goes

[astral comet]

Whatsup guys, i like using the "create" tab on civit.ai, which allows you to generate images with a limited set of models. It only does low res, so when I get something I like, I just copy the info and seed in auto1111 with the same model, then upscale it in img2img. It used to work perfectly, but recently when I do that it doesn't create the image I got in citiv.ai.
I went through all the settings: model version, prompts, negative prompts, sampler, steps, clip skip, etc. And theyre all the same. Anyone has a clue what I could be overlooking?

[wise flame]

how to fix this error?

ImportError: cannot import name '_TORCH_GREATER_EQUAL_1_12' from 'torchmetrics.utilities.imports

[ornate elk]

that tool (A1111 web ui Autoinstaller) is outdated

you should check my SD install guide in the Pinned Messages of this Channel

If thats an old installation you need to delete the venv folder

[wise flame]

thanks

[hoary trout]

Welp
Didn't help
I know for a fact that it's ram use causing it now tho

Was monitoring ram usage and it ran right up to 99%

[wise flame]

100% working, thank you so much

[ornate elk]

no problem🙂

[ornate elk]

you could try to increase the pagefile for windows, but your laptops uses a slow hdd so it wont help much in terms of speed, try lower the resolution or the stuff your doing in SD

[hoary trout]

i'm using a desktop

[late temple]

for some reason my PNG Info just doesnt work despite working on the same ing for others. Now i cant even look up some img from Civit witch are png and have all info on them.

[ornate elk]

Oh okay. Must be an old one then.
Wouldn't recommend using SD only on CPU if it takes an hour for an image.
You could try the StableHorde client (free) or Leonardo.ai (daily free limit) for example to generate images.

[ornate elk]

Make sure the webui is whitelisted in any browser adblocker

Also make sure your webui is updated

[hoary trout]

do they allow nsfw?

[ornate elk]

The stablehorde client allows nsfw

Idk about Leonardo (but I dont think it does)

[hoary trout]

please tell me that's something i can just open up on my browser, and not something else i have to install

[ornate elk]

Yes they have some browser bots:
https://tinybots.net/artbot
And https://artificial-art.eu/
And a simple .exe Programm downloadable from here:
https://dbzer0.itch.io/lucid-creations

More information on StableHorde:
https://stablehorde.net/

[tulip furnace]

henlo guys. probably asked to oblivion but i cant seem to find a solution. why does my controlnet seem to not affect anything or just simply produce weird images where the stick figures show up in the results. edit: nvm, figured out preprocessor and model both have to be canny💩

[supple fable]

MY SD XL looks perfect until it finishes, then it looks like this.. What gives

Is it cuz of loras?

Im using realvis 3 XL version with automatic VAE

[onyx solar]

That means the model you have likely doesn't have a baked-in VAE and you need to use the standard SDXL VAE.

[supple fable]

Im using this one.. says its baked in.. https://civitai.com/models/139562/realvisxl-v20 Im tryna find best XL model, dont even know difference between regular and XL

do loras not work on XL?

[onyx solar]

LoRAs work fine with XL as long as it's a LoRA made for SDXL. However, the result you're showing is a VAE issue. Try switching the VAE temporarily and then back to Automatic.

[supple fable]

Ah I fixed it. Doing exactly that fixed it. Whats the biggest difference between XL and regular?

[onyx solar]

Well, 1.5 isn't "regular", just an earlier version. The largest difference is that natively, 1.5 works best on ¼ megapixel (512x512) image sizes and SDXL is geared toward 1 megapixel (1024x1024) sizes. Depending upon the model, it's also likely to follow your prompts better without (or with fewer) negatives and reacts better to more natural language positive prompts (though that's not necessarily always true).

[supple fable]

Huh. Interesting

I was playing with online generators and said.. this is dumb, I can probably run it natively, too bad Idk what various sites uses for models

[onyx solar]

Depends on the site. Some have their own custom models in use. But go to Civitai or Tensor.art and look around.

[supple fable]

I been using realistic vision, I just wish I could make it do more.. fun stuff without loras lol 😛

[onyx solar]

Most models are pretty good on XL and can do pretty amazing stuff with the right prompts.

[supple fable]

😮 hmm..Like, im typing in poses, and they're just mid waist shots ignoring pose lol

[onyx solar]

Don't think so much about the pose first...think about describing something you want to see in the image, then worry about the pose. If you want to see full body, describe something like the shoes AND the hair.

[supple fable]

Ah! Ok 🙂

[undone badge]

Hey, I hate that my first post here is going to be reaching out for help but I haven't been able to troubleshoot my issue on my own. I'm following along with a guide on installing AUTOMATIC1111 on Arch Linux with an AMD card. When I use webui to launch it I get what I believe to be an error followed by it crashing. ./webui.sh: line 256: 130981 Segmentation fault (core dumped) "${python_cmd}" -u "${LAUNCH_SCRIPT}" "$@" as while it does launch in my browser I get the following error "Connection errored out." Any advice on how to resolve this issue would be appreciated and thank you in advance.

Attached the full output in my terminal, my webui-user.sh file, and the guide I'm using for added context.
https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Install-and-Run-on-AMD-GPUs#install-on-amd-and-arch-linux

[unique glacier]

venv "H:\A1111 new\stable-diffusion-webui\venv\Scripts\Python.exe"
No Python at '"C:\Users\shawal\AppData\Local\Programs\Python\Python310\python.exe'
Press any key to continue . . .

i am using python i downloaded from microsoft store. it appear that python is not in path. How do i add it? i tried to open python file location but it wont allow me. it seems microsoft store obfuscate where python is install.

i cant use python installer as it seems that it is not install properly as i had trouble with it. when i use python from microsoft store it work fine. idk why

[narrow kestrel]

121 hours, what have i done

[valid gorge]

RTX 3060 12GB, it's currently on sale and would be perfect for my budget. Gettin 4060TI 16GB would be great but at the same time it's 40% more expensive. I'm not planning on doing training or anything, just plain t2i and i2i and upscaling. Is the 12GB enough for my needs?

[ornate elk]

You need to install python 3.10.11 64bit to make it work

Select "add python to path" when installing

[ornate elk]

Yes would be enough for generating highres images and your also be able to train with it 1.5 models and loras

[fair oxide]

I followed the instructions. It didn't work. It got a bunch of errors during installation.

[ornate elk]

Okay do you have the cmd log with the errors?

[fair oxide]

Yup.

[valid gorge]

Thanks! Maybe I just should go with 3060, still wondering would the 40% more in price of 4060TI be worth the difference. Maybe not, unless I want to start training models in the future.

[ornate elk]

Feel free to share it here, I can take a look

[ornate elk]

More vram is always better. The problem with the 4060ti is the lowered bandwith.
But it still has more power than a 3060.
Also we dont know what ai stuff needs in the future. But for now 12gb can do everything really well.

[valid gorge]

Very good pointers! Maybe getting the 3060 12GB for now will satisfy my needs and in time getting a second hand 3090 24GB would be the most ideal. As far I understand 3060 vs 4060TI, the latter is about 20-25% faster. Maybe not worth the extra 40% in the price tag.

I'm currently using a Mac (M1 MAX) which is reasonably ok but not the fastest. Takes around a 1:30 minutes to create a single 1024x1024. I spend most of my time inpainting, so it means a lot of waiting and it's hard to batch things.

SDXL text2image, with 2x hires.fix with basic settings, 700x1000px --> 1400x2000 take around 10+ minutes for on my current computer.

[ornate elk]

For inpainting make sure to set the Inpaint area to "mask only"
And then you can lower the resolution to 768x768

Should be faster

But yea 3060 is a good card for SD

3090 is rare but still a beast with 24gb vram

[valid gorge]

It's a bit faster when inpainting with 768, I just want all the details haha (not sure if 1024 vs. 768 is a big dirrence on that regard haha)

[valid gorge]

And still quite pricey even second hand. I'm guessing miners keep the prices up for second hand cards

[ornate elk]

If your inpainting with sdxl images you should use 768 or 1024, if you inpaint with 1.5 models then you should use 512x512

[valid gorge]

Yep 👍

[ornate elk]

thats a connection issue. Make sure to use ethernet (or good wifi), disable vpn (if used), restart the PC, delete the venv folder and relaunch the webui-user.bat

[fair oxide]

I don't have VPN and I don't use WiFi. The internet works fine.

[ornate elk]

but you downloaded with 500kb and then the connection errored out.
So the best you can try is:
open up a cmd and type
pip cache purge
then delete the venv folder
then launch the webui-user.bat

[fair oxide]

I already did that before installing, according to your instructions.

[ornate elk]

yes, but to fix the broken files you need to do it again

[fair oxide]

Ok. Doing it now.

[valid gorge]

Sometimes A1111 outputs a gray imge, when the creating process is going the image looks fine, after finishing it turns weird and gray. Is this a bug, memory issue or something else?

[ornate elk]

thats strange, what are your commandline_args=?

[valid gorge]

--no-half --upcast-sampling --no-gradio-queue

--skip-torch-cuda-test --upcast-sampling --no-half-vae --use-cpu interrogate

[fair oxide]

It worked! Thank you so much!

How do I launch Stable Diffusion? What's the launch icon?

[ornate elk]

you can try these args for Mac: --skip-torch-cuda-test --opt-sub-quad-attention --upcast-sampling --no-half-vae --medvram-sdxl --use-cpu interrogate
its from the Mac install guide from Viking:
https://github.com/viking1304/a1111-setup/discussions/2

[ornate elk]

nice no problem 🙂
to launch SD you always need to run the webui-user.bat

[fair oxide]

Wait, no it didn't. I am confused. CMD says that it failed to load, but it opened the browser and it seems to be working.

[ornate elk]

you can make a desktop shortcut for it

[valid gorge]

Thanks, will test those!

[ornate elk]

you need to leave the cmd (black window) open after launching it

thats the actual programm in the background

[fair oxide]

It is open. I didn't close it. Here's what it says.

[ornate elk]

oh okay, i see you have a typing error in the webui-user.bat
you need to use --lowvram but you have typed --lowram

[fair oxide]

I tried to generate a basic image and after 11 minutes of processing got this: RuntimeError: Could not allocate tensor with 419430400 bytes. There is not enough GPU video memory available!
Time taken: 11 min. 45.2 sec. A: 3.03 GB, R: 3.88 GB, Sys: 3.9/3.88086 GB (100.0%)

[ornate elk]

what was the resolution?

[ornate elk]

also 4gb vram is very low for the AMD webui.
the base model is 4gb so thats why its so slow. also the model is not good.
the best you can do is to use 2gb models like Dreamshaper v8 for example:
you can get it here: https://civitai.com/models/4384/dreamshaper
After you downloaded it you need to put it into the models/stable-diffusion folder.
Then restart the webui
and for the resolution try 512x512 or 512x768 for portrait format

[grand fable]

Hi, does any of you have some experiences with running ComfyUI on RunPod? I would need some help with initial setup and installation..

[lost crescent]

Where'd the SDXL Room go to? It's gone from my Discord!

Why does A1111 have to take a month of Sundays building yaml files every time u change a model?!

[lost crescent]

https://github.com/FurkanGozukara/Stable-Diffusion/blob/main/Tutorials/How-To-Use-ComfyUI-On-Your-PC-On-RunPod-On-Colab-With-SDXL.md

https://blog.runpod.io/how-to-get-stable-diffusion-set-up-with-comfyui-on-runpod/

https://github.com/blib-la/runpod-worker-comfy

[valid gorge]

Sweeeet, by adding --opt-sub-quad-attention and --medvram-sdxl sped up things almost 30% or me! Thanks man!

[ornate elk]

awesome np 🙂 thanks goes to Viking, as he tested and compared them all on his Mac

[grand fable]

I found this one.. I'm more interested with some experiences for example what type of storage (normal / network) to use, what size is ok etc. 😉 but thanks anyway 🙂

[mystic quail]

Hello, I've just installed the "Riffusion" extention and now I'm running into this error, how do I fix it?

[ornate elk]

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

[mystic quail]

I've got an Nvidia RTX 3070

My webui-user.bat file is:
@echo off

set PYTHON=C:\Users\AppData\Local\Programs\Python\Python310\python
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --xformers
call webui.bat

This has only started happening after I installed that extension.

[ornate elk]

yea seems like your webui is to outdated for the Extension. your on 1.4.1 and the latest webui is 1.7.0

[mystic quail]

oh wtf

jeez

ok

Stupid question

how does one go about updatign

[ornate elk]

to fix that you need to add:
git pull
to the webui-user.bat in an empty line before the webui.bat

but also add --medvram-sdxl --no-half-vae to the commandline_args=

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

[mystic quail]

sorry where do I add "git pull"

[ornate elk]

edit the webui-user.bat and replace everything with that:
`@echo off

set PYTHON=C:\Users\AppData\Local\Programs\Python\Python310\python
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --xformers --medvram-sdxl --no-half-vae
git pull
call webui.bat`

then save the bat, and delete the venv folder (important)

then launch the webui-user.bat

[mystic quail]

does this update user.bat to 1.7.0?

[ornate elk]

this updates the whole webui to 1.7.0

[mystic quail]

oh ok thank you so much

how do I update sd because I think im still running 1.4

isnt the latest like 1.5

[ornate elk]

when the updates is done and the webui opens again, scroll to the bottom in browser to check the version.
also you need to update your extensions in the extensions tab

[mystic quail]

alr

[ornate elk]

with that your mostly mean the checkpoint or model your using right?

[mystic quail]

Yes i think so

[ornate elk]

you should use 1.5 or SDXL based models.
1.5 based models are trained on 512x512 resolution and good supported by extensions and the community.
SDXL based models are trained on 1024x1024 resolution, using more vram, and are slower but highre res.
You find a ton of models you can download and use on civitai.com for free.
models(checkpoints) are larger than 1.98gb and go into the models/stable-diffusion folder to "install" them

[mystic quail]

Do I need to install anything else to use a SDXL model?

[ornate elk]

nope

as you added --medvram-sdxl --no-half-vae to the webui-user.bat your ready to use sdxl models with a good performance

[mystic quail]

oh ok

even with my 3070?

[ornate elk]

yes

worked for my 1080 too

[mystic quail]

alr thanks

Ive now run into this issue

[ornate elk]

a full screenshot would help

[mystic quail]

oh wtf

i thought i did sorry

[ornate elk]

np ^^

[ornate elk]

with full i meant everything from the top

[dusty meteor]

unfortunately SD WebUI has stopped working for me again since yesterday 😦 And I don't understand why. would appreciate any help you could provide. https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/14761

[dusty meteor]

does anyone know what -[MTLIOAccelCommandBuffer validate]:207: failed assertion' commit an already committed command buffer' means?

[ornate elk]

Okay to fix that:
Close the cmd.
open up a cmd and run
pip cache purge
After that close the cmd.
Go into the **stable-diffusion-webui **folder.
Click in the File Explorer bar. (not in the search bar).
then type cmd and hit enter.
after that type and run:
git reset --hard HEAD
hit enter and then
git pull
Then delete the venv folder and run the webui-user.bat

[mystic quail]

its installing stuff, ill update when its done

[mystic quail]

It has been stuck at his for quite the long time

nevermind alas it has continued

[fair oxide]

have been having an issue where when i try changing the checkpoint and it doesnt change and it says like 0.3 seconds til it changes to the checkpoint but then i waited til 200 seconds and realized that 100 percent of my ram was being used

idk whats going on

so i've just been stuck on the one checkpoint

cause my ram is normal til trying to change checkpoints

[ornate elk]

hey, whats your GPU and whats inside your webui-user.bat?
also how much RAM?

[fair oxide]

i have a 16 gb t force ram and 8 gb 3070 i have in my batch file --xformers --autolaunch --medvram --upcast-sampling

[ornate elk]

okay, change --medvram to --medvram-sdxl and --upcast-sampling to --no-half-vae

also how much free space do you have left on the C: drive?

[fair oxide]

ok im gonna try also could it also be that the person helped me had me add other things to my ui

76

gbs

[ornate elk]

idk, (but i dont think so) but make sure other programms are closed when using SD

if your not sure if the person that helped you was legit. then you can also do a reinstall with my guide in the Pinned Messages of this Channel

[fair oxide]

this is my setup

also sadly its still doing it as you can see

in the top left

[ornate elk]

can you try to load a smaller model?

[fair oxide]

i gotta find one

[ornate elk]

https://civitai.com/models/4384/dreamshaper

that one is good to test

only 2gb

[fair oxide]

if it doesnt work i might reinstall it

i will say

when i changed my medvram

to the sdxl

it wasnt taking up 100 percent of my ram

and was only using up just a good portion

so that part was fixed

[ornate elk]

thats good

[fair oxide]

i like civitai i just wish people stopped focusing on the porn aspect of the ai and the space and cosmic horror aspect

i love making cosmic horror images

[ornate elk]

your webui looks normal, but the model your trying (moistmix) is 7gb in size and also old

so i would recommend trying other models or delete the 7gb one

[fair oxide]

yeah but when i used to have this ai a while back i had like 10 gb models that loaded in 10 seconds

so idk why this is happening

[ornate elk]

if models still wont load you may need to adjust your windows pagefile

[fair oxide]

could it be my ai is on my d drive and not on my c drive

[ornate elk]

if your D drive is a slow hdd than that can be an issue

[fair oxide]

yeah it is hdd

seagate

[ornate elk]

Laptop?

[fair oxide]

im on pc

[ornate elk]

ok

[fair oxide]

custom built

when i first was building it the fans werent working and my cpu was going up to 90 to 100 c in like 2 seconds

but i found out what was wrong

[ornate elk]

ohh xD

[fair oxide]

gonna test it now

just imported the checpoint

my wifi a bit slow

[ornate elk]

okay

[dusty meteor]

solved this. nvm

[fair oxide]

turns out i think it might be the ai its self and not just checkpoints

i tried changing the SD VAE

and its doin the same thin

[ornate elk]

okay, then make sure all adblocker aswell as opera vpn are disabled or whitelisted the for the webui

[fair oxide]

ok im trying it now

thanks thats a good point to amke

maybe its saying processing

and its just glitched

ima just type a random prompt

and see if it generates

cause it didnt fix

yeah it just says in queue:1/1

[ornate elk]

what does the cmd shows?

[fair oxide]

thats what i just screenshotted lol

more up close

[ornate elk]

and you pressed generate?

[fair oxide]

yeah

[ornate elk]

then there is still an adblocker blocking

[fair oxide]

is there a way i can maybe change from opera gx

cause im pretty sure

everything is disabled

oh yeah i gotta change default

default website

[ornate elk]

install firefox for example

[fair oxide]

installing now

[ornate elk]

also use an other adblocker, for example Ublock Origin

much better for privacy and performance and its opensource

[fair oxide]

i set firefox as default browser

now what

[ornate elk]

open the webui in it

[fair oxide]

ok

my friend told me i should use firefox instead of opera

i guess its just better?

[ornate elk]

yes, less bloated and better for privacy aswell

[fair oxide]

i thought it was working for a second cause this one said 6.0s seconds and it started looking like it was working

hold on let me get a clip without my spotify music playing to show you what i mean

nevermind it didnt do it again

it just doing what it did last time

im gonna completely retry installing

this and see if it works

and be honest do i really need the extra SD-VAE stuff in my ui cause last time i was using this i didnt have that stuff up there just the checkpoints

and everything was chillin

OH SHIT

WHAT

it workin

yooooo

nevermind

i just kept spamming the reset button

and then it brought me back to default checkpoint

then when i went to dream shaper

i told me it needed 20 seconds and it actually took 20 seconds

its workin thanks

opera is just boof

[ornate elk]

go into settings - user interface - quicksettings
there add: sd_vae
then hit apply & reload ui

[fair oxide]

MY CAT

stepped on my power button

[ornate elk]

RIP

[fair oxide]

while im reopening the ai

how caught up on one piece are you

[ornate elk]

next pc case with a sidemounted powerbutton 😛

[fair oxide]

fr cause shes stepped on my power button in the middle of a tf2 game

it was tour night

this is what my quicksettings look like

sd_vae was already added

[ornate elk]

then you should see the dropdown

in txt2img etc

[fair oxide]

reloading ui

oh yeah i do

thats cool

i dont remember that being a thing

for text 2 image

is this new

[ornate elk]

nope

[fair oxide]

nvm im looking at img2img

mb

im moving to txt2img

whats your recommened sampling steps

i was told 30

[ornate elk]

yea 30 is good

[tame meteor]

I have a question. I tried to train LORA on my girlfriend's photos. I used Kohya LoRA training with 100 photos of her but the results turning out like that. It definetly tried to match the face, and it kinda did, but it looks nothing like I want it to be. I'm using ChilloutMix as training model. Could somebody help me?

[tame meteor]

tweaked some settings, yet it still doesnt look good

[summer onyx]

Is there anyone who can help me fix a technical issue with training a LoRA using Kohya?
During the Kohya install I chose to install bitsandbytes and selected the latest version. I am now running a Lora Training and it is failing when trying to use bitsandbytes
File "C:\Users\Kevin\Kohya\kohya_ss\library\train_util.py", line 3497, in get_optimizer
import bitsandbytes as bnb
ModuleNotFoundError: No module named 'bitsandbytes.cuda_setup.paths'
Did i install it in the wrong place or install the wrong version?

[willow oracle]

hi there! first time on this channel - trying the Sketch to image API for a class using javascript...but keep getting 403 error: Invalid API key. I tried generating multiple but with the same issue...any reason why?

[summer onyx]

I found a workaround on stack overflow:
*I've also met the same problem. the way i solve it is by changing the optimizer to the non-8bit version. like if you have selected AdamW8bit, you change it to AdamW this time, then the training will work with no problem. *
Its running now and will take 1hr 17min so I'm crossing my fingers. 🤞 If anyone knows how to fix it that would be good as i would prefer to run the 8 bit version.

[sonic ferry]

Can someone tell what I'm doing wrong here? I just added the Controlnet for hands, now I'm not sure how to continue it through the Upscaler below before going to the next KSampler.

The MeshGraphormer takes an image, but that image isn't upscaled yet, so I was thinking I either have to upscale the image first, out of the Upscaler and also the latent out to the Set Latent Noise Mask before that Latent goes to the KSampler...that sound right? :/ Edit: No, can't have 2 Latents going into the KSampler...

[languid rock]

Hello! super new to all of this, is this an ok place to have questions on the functionality of SD? for instance i'm learning how to "train" and cannot find the "preprocess images" tab

[mystic quail]

sdXL generation taking 20 mins to create an image 512x768!? is this normal or can i fix this

[sonic ferry]

I moved my Upscale to a third KSampler, but kept the same settings from the KSampler before it...What should I change? My image is all pixelated, I thought it was an issue from the ControlNet (which I'm now not sure I'm liking).

[tame meteor]

what's your hardware

[mystic quail]

I've got an Nvidia RTX 3070, with 32 gigabytes of ram, with a AMD Ryzen 5 3600

[tame meteor]

you're using automatic1111 or comfyui?

[mystic quail]

a1111

[tame meteor]

change to comfy👍

i couldnt even run sdxl on a1111 but on comfyui i can generate sdxl in less than minute on 2060

[mystic quail]

How do I change?

am I still able to use all my other models

and loras?

[tame meteor]

yes, of course

[mystic quail]

alr

[tame meteor]

https://github.com/comfyanonymous/ComfyUI

it might look confusing first time but its really simple

[mystic quail]

how can you upscale in comfyui

[tame meteor]

with upscale model?

[sonic ferry]

Lot's of different ways, this is what I'm doing.

[tame meteor]

i think that's the simpliest way

just putting upscale image between vae decode and saveimage

[sonic ferry]

Here, take this. https://civitai.com/models/216482/tangoone-sd15-workflow-with-upscales Easy enough to get you started.

[tame meteor]

yeah, and that looks complicated haha

[sonic ferry]

Oh, well I've complicated my own a lot more since I found that the other day. LOL

[tame meteor]

i'll just stick to my workflow for now haha

[sonic ferry]

https://tenor.com/view/im-fine-little-boy-crying-gif-15238441

[tame meteor]

"comfy" lol

[sonic ferry]

You should try the Efficiency Nodes, would be so simple with your setup.

[tame meteor]

i'd be afraid to change anything here

[tame meteor]

What's that?

[sonic ferry]

LoRAs on the left then Checkpoint, Latent Image, Prompts, VAE Option all together in the second, a KSampler I find better, and right to Save Image, skip the VAE Decode...Unless you need it for another process.

Ah, this Efficiecy Loader has the spot for a LoRA I think is mainly used for LCM.

Could probably use that if you're just using a single LoRA too.

[tame meteor]

I will check that out after I deal with making LoRA

[mystic basin]

is this normal? i still have a lot of vram left

sorry that my windows is in french but in english

it says 12.5/19.6 go for graphics processor memory, 11,0/12,0 go for dedicated graphics processor memory and 1,5/7,6 go for shared graphics processor memory

the utilization of my gpu is at 0% right now because i'm not using it but when i am, it goes to around 99%

i don't understand why, all the process goes well but when like the image is almost done at like 97%, i get a runtime error

[ornate elk]

Can you show your txt2img settings?
It's not normal.
Your card should be taking 1 minute Max for an sdxl image

[ornate elk]

What are your txt2img settings when getting this?

[mystic basin]

either this or

this

it happen with these two setting

[mystic quail]

wdym by settings? steps, sampler, dimensions or something else?

[ornate elk]

Yes

Hires fix if used

[mystic quail]

I didnt use that

I was using 30 steps, euler a, 512x768, cfg was 7

[ornate elk]

That resolution is to high. You need to use lower resolution and then use hires fix or SD upscale script in img2img to upscale

[ornate elk]

But that should take less than a minute for your card

Make sure to launch the webui-user.bat

Not webui.bat

[mystic quail]

this is the one i launch

[ornate elk]

Then make a screenshot if you can reproduce the issue

[mystic quail]

yeah thats what im doing rn

[mystic basin]

i tried using 960 by 540 and use the hires option to upscale the image by 2 then press generate but it still happen

[mystic quail]

Could this have anything to do with it? i've only just got this

[ornate elk]

First thing to know for AMD:
When you get the out of memory error you always have to restart the webui completely to get the vram back.
Then pls try with these settings:
512x768
Steps: 30
Hires steps: 10
Upscale by 2
Denois 0.5
Upscaler: Resrgan4x+

[ornate elk]

Yes thats the wrong python version

And can cause these issues

[mystic basin]

alright i'll try

[mystic quail]

oh

how does one go about updating that

[ornate elk]

The latest supported version is 3.10.11

@mystic quail
Uninstall the 3.11 if you dont need it.
Then install 3.10.11 64bit, and checkbox "add python to path" at install.
Then delete the venv folder
Launch the webui-user.bat

[ornate elk]

Make sure you dont have Wallpaper Engine or epic games launcher running in the background. To save vram usage.

[mystic basin]

alright

[ornate elk]

Let me know if you get the error again

[mystic basin]

i just realized that i don't have xformers

[ornate elk]

Thats okay, as xformers is for nvidia only

[mystic basin]

oh ok nice

i'm testing the setting

[mystic quail]

In my control panel it says im running Python 3.10.6

[ornate elk]

Okay that will work too

[tame meteor]

why does everyone use a1111, it's totally not worth to use if you don't have 4090 lol

[ornate elk]

You still need to delete the venv after uninstalling 3.11

[mystic quail]

how do i uninstall 3.11?

[ornate elk]

Why? Its an easy to use webui

[ornate elk]

In the system control

[mystic basin]

it didn't work

[tame meteor]

amount of problems with it makes comfyui so much better

[mystic quail]

im confused

it says 3.10.6

not 3.11

[ornate elk]

Yea maybe you installed it trough the winows store

Check the apps in the new windows settings

[mystic basin]

you can do that in the windows setting. sorry for the french

to delete the other version of python

[mystic quail]

yes found it

thank you

[ornate elk]

What problems did you encounter?
For the most people auto1111 works fine.

@mystic basin
Okay we can make upscale work as I did on a 6700xt before.
You need to install the tiled diffusion extension first

[tame meteor]

couldn't even run SDXL, long generating times, problems with loading loras

[hallow edge]

can anyone help me with installing kohya_ss on windows 11? I got all the required things installed before I started, then rand into an issue with ".\setup.bat".

First issue was because I had more than 1 version of python installed so I uninstalled all my python versions and reinstalled 3.10.9

Then I ran it and It told me I had to disable python in the app execution aliases so I did that.

But now after that, I get no errors... but I also get absolutely nothing back

[mystic basin]

alright i'll install tiled diffusion

[ornate elk]

Let me know when your ready then I'll give you the test settings

[mystic basin]

this is the extension right?

[ornate elk]

Yes

[mystic basin]

now i should have it

[mystic quail]

I now get this

[mystic basin]

i have it

[ornate elk]

Make sure you restarted too.
Then try the following settings:
512x768
30 steps
Hires fix settings:
Hires steps: 10
Denois 0.5
Upscale by 2
Upscaler Resrgan4x+

Tiled VAE: On let everything default but change:
Encoder Size to 1024
Encoder Size to 128

Dont use Tiled Diffusion. Only tiled vae for upscaling with hires fix

[mystic basin]

alright i'll try

[ornate elk]

You need to install python again

Either 3.10.6 or 3.10.11 and checking add python to path

[mystic quail]

but dont i already have it installed

ok

[mystic basin]

sorry gtg eat, i'll come back after to see if it work

[mystic quail]

its installed Python 3.10.11

how do I check add python to path

[ornate elk]

In the python installer

At the first window after clicking the setup

[mystic quail]

i installed through microsoft store

there was no setup

[ornate elk]

.....

That won't work

[mystic quail]

oh

[ornate elk]

Uninstall it again xD

[ornate elk]

https://www.python.org/downloads/release/python-31011/

At the bottom the 64bit installer

[mystic quail]

thank you

does it matter whether it is 3.10.11 or 3.10.6?

[ornate elk]

Nope

But 3.10.11 has more bugs fixed

[mystic quail]

ive installed it but im still getting the same error message

[mystic basin]

can't reinstalling a1111 entirely work?

i'm back

the thing is not detecting the other python version, reinstalling a1111 could probably work

[ornate elk]

Did you checked "add python to path"?

[mystic quail]

i dont think i got an option to

[ornate elk]

Okay then run the installer again.
Select modify, next, then check "add python to system variables"

[mystic basin]

i can't find tiled VAE. i 'm not sure if it's my eye that are the problem or it's because i didn't install it correctly

[ornate elk]

Scroll down

In txt2img

It should be there

[mystic quail]

when i run it, do i click install now or customize installation?

[mystic basin]

it's not there, it's me that didn't install it correctly

[ornate elk]

Customise

[ornate elk]

Yes probably

Best you can do is to delete it from the Extensions folder and then install it correctly with the in build extension installer inside the webui

[mystic basin]

this is the last thing i see

[mystic basin]

i'll try that 👍

[mystic quail]

Im still getting the same error? ive checked add python to path

[ornate elk]

Click on extension tab. Click on available. Click on load from

In the list search for tiled

[mystic basin]

alright

[ornate elk]

Okay is there a venv folder to delete?

[mystic quail]

no 😞

[ornate elk]

Ah damn.

[mystic basin]

i found it

i think that if you reinstall a1111, it will work

[ornate elk]

No it wont

[mystic basin]

oh i though it would

[mystic quail]

is doing a good ol' restart the computer worth it?

[ornate elk]

Open up a cmd.
Type
Where python
Copy the path it gives you.

Edit the webui-user.bat
At the line Python=
Paste the path.
Save and launch

Make sure you copy the path that says /appdata/

Not the Microsoft store stuff

[mystic quail]

This one? C:\Users-------\AppData\Local\Programs\Python\Python310\python.exe

[ornate elk]

Yesssss

[mystic quail]

ok

[mystic quail]

where in the webui-user.bat do I add this

[ornate elk]

At the line
Python=

[mystic quail]

@echo off

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

call webui.bat

[ornate elk]

set PYTHON=

What was your GPU again?

[mystic quail]

Nvidia RTX 3070

[mystic basin]

it worked, thanks for the help

[ornate elk]

No problem. With these settings you can also do 540x960 for FullHD

[ornate elk]

Then at the line COMMANDLINE_ARGS=
Add: --xformers --medvram-sdxl --no-half-vae

[mystic basin]

W community guide

[solid wind]

Hey

[mystic quail]

@echo off

set PYTHON=C:\Users\Melanie\AppData\Local\Programs\Python\Python310\python.exe
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --xformers --medvram-sdxl --no-half-vae

call webui.bat

Like this?

[ornate elk]

Perfect

Now everything should work

[ornate elk]

Hey

[mystic quail]

its creating a new venv folder now (yay)

[ornate elk]

No problem, but you can't Upscale sdxl with hires fix with these settings. For that or other images that go over 1500pixels you need to use the SD upscale script in img2img.
Let me know when you want to try it

[torn spear]

I'm using this checkpoint, and it keeps telling me error. " Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check." Can someone help me with this ?

[mystic basin]

thanks for the info, i do not have any sdxl model at the moment. but for now i am fine without any so when i decided to use one i'll do that

[ornate elk]

Alright 🙂 👍 for any other amd related questions in SD feel free to ask me

[mystic basin]

👍

[ornate elk]

Hey, how big is the checkpoint file?

[torn spear]

Oh right

Let me check

It's in the MB. I forgot 2GB or higher

[ornate elk]

Yea then its mostly a lora

Thats why you get that error

Loading loras as models won't work

[sonic ferry]

Can anyone help me with some ComfyUI mess?

[mystic quail]

might be a silly question

but can sd run alright on an hhd? as apose to an ssd?

or can it run at all?

[ornate elk]

Yes it can run

[mystic quail]

does it effect performance

[ornate elk]

Not when generating

[mystic quail]

ok

[ornate elk]

As there it only uses gpu

[mystic quail]

alr

[ornate elk]

But for loading models and saving images

It takes more time

[mystic quail]

how could i go about moving it over? can I just copy the stable-diffusion-webui folder over?

[mystic quail]

oh ok

[ornate elk]

Moving it will break the auto update feature.
But that can be fixed

[mystic quail]

so i can literally drag and drop the whole folder?

[ornate elk]

Yes

Here is my copy paste fix after moving it.

@mystic quail
As you moved your webui you won't be able to update it anymore.
To fix this go into the stable-diffusion-webui folder and click in the File Explorer bar (not searchbar) and type cmd and hit enter.

Then use the following commands:
git init git remote add origin https://github.com/AUTOMATIC1111/stable-diffusion-webui.git git fetch --all git reset --hard origin/master git branch --set-upstream-to=origin/master master git pull
Then launch the webui-user.bat

I'm off for now, its getting late

[mystic quail]

alr tysm

[ornate elk]

No problem 🙂

[proper peak]

how can I get stable difusion to make these but in another style ?

or something similar

becaise

[tame meteor]

Why does lora ALWAYS put same person on shirt😭😭😭

can somebody help?

[sonic ferry]

Skip to 11 minutes, I think that's what you're looking for. https://www.youtube.com/watch?v=_C7kR2TFIX0&t=665s

[tame meteor]

Okay i get it, but even without anything it still tries to make 2 faces

I will check if putting "2 faces" on negative prompt will make work

[torn spear]

The Lora files I am putting in the Lora folder isn't showing on Stable Diffusion ? Can someone help me on this

[onyx solar]

Do you have an SDXL model chosen and are looking for SD 1.5 LoRAs or vice-versa? You will only see LoRAs applicable to the model you have selected at the time. Also, if you didn't refresh the list of LoRAs or restart whatever you're using, you won't see them.

[torn spear]

Oh, let me check which one is it

I think the base model is other

[onyx solar]

Model = Checkpoint

What version of Stable Diffusion checkpoint are you currently set to?

[torn spear]

Nvm, I figured it out

Thank you though

[onyx solar]

What was it?

Don't be afraid to share the solution, even if it's for something simple. There are new people that join every day that repeat the same mistakes others make and listing how you fixed your own issue helps them out when they do the same thing.

[torn spear]

I think for me I needed SD 1.5 model

The other ones were not the ones I am looking for

[primal star]

hii would anyone know how to solve this problem in webui-user.bat file? it always stops on 'launching webui with arguments" and then "press any key to continue.." then when i press a key, the whole command prompt exits:(

i have nvidia geforce rtx 3050

[sharp robin]

What are the easiest ways to make images on AMD GPUs? WebGPU?

[woeful sedge]

hello, why i dont have preprocessing image/model on the train tab

[ornate elk]

Thats a question for #📝｜prompting-help

[ornate elk]

The easiest way is to follow my install guide for AMD. Its in the pinned messages of this channel.

[ornate elk]

Hey, remove --skip-torch-cuda-test --precision full --no-half --lowvram

Instead add: --xformers --medvram --no-half-vae
Then save.
Then delete the venv folder and relaunch the webui-user.bat

[tender cosmos]

)))$$

[dusky wind]

Hi! why can't I send promt to the bot?

[ornate elk]

Hey, its down for maintenance:
#1047610792226340935 message

[agile bobcat]

Greetings.

There is a constant problem with dreamstudio api. There is censuring of what does not need to be censored. There are a lot of complaints, I don’t want to switch to another service.

What can be done?

Here's an example. Prompt: girl, brunette, brown eyes, necklace in the shape of a dragon. short shoulder-length hair, danger, shards of glass, shards of a mirror, moonbeam, background, fog, cold look, full of hatred, hair, pose: looking at the viewer: gloomy look, looking into the soul, black hair, long blue-pink dress with sleeves with black rose print, moonlight, red lips, black shadow, fog, moonlight on background, candles, paintings,

[vocal burrow]

I would contact stability.ai staff directly through mail to discuss that kind of dreamstudio problem.

In the meantime you can always make use of the negative prompt.

(or use a local instance but that might not suit your needs if you initially went the online route)

@agile bobcat

[agile bobcat]

We use the online option, since the large flow of users and our capacity do not limit the load(

Let's try to configure negative prompts

Thx

[astral comet]

If i train a lora with only photos of a head, will it only affect the head and not affect the base model for the body?

[paper ocean]

where can I get the stable dreamer bot model so I can use it locally? would I be able to generate the videos if I host it or just images?

[elder turtle]

The model the bot is using right now isnt public I think, but you can run SDXL and its custom models locally for example

[paper ocean]

yeah I've been trying to do this but seems like my 6gb gpu is not enough, do you happen to know any paid services that would allow to have a similar workflow to this bot? I've tried runway but the videos ere really inconsistent, my boss really wants the style of video that the bot was generationg so i guess asking him to pay for a platform is not a problem

[valid mulch]

Anyone else here using a 3060ti? If so, what are your command line args for Automatic1111? I'm wondering if I'm properly optimizing... keep ending up with black images all of a sudden.

[ornate elk]

A 3060ti only needs:
--xformers --medvram-sdxl --no-half-vae

[valid mulch]

Thanks! I think that may be what I already had. I'll have to check.

[ornate elk]

If you still get black images, check the model size and make sure its larger than 1.98gb

[valid mulch]

I think I was able to solve the black images by just rendering a ton of 64x64 images with Euler A. Seems to have 'flushed the pipes' as it were...

[ornate elk]

Okay still strange xD

[valid mulch]

Yeah. I'm just starting to meddle with Deforum and all of a sudden all sorts of things are acting weird. lol

[fair oxide]

im using rtx 3060 and only use --xformers w/o any issues, you sure you need --medvram-sdxl --no-half-vae for 3060ti?

oooh, i see 3060ti has 8gb vram while 3060 has 12gb

[ornate elk]

thats why --medvram-sdxl is needed 🙂

[fair oxide]

yes, i understand that now, cause i was mistaking 3060ti for 12gb vram

[ornate elk]

yea that was an nvidia prank on the community

give the faster card less vram...

[fair oxide]

interesting they did that, but does ti version has some other enahancements?

The RTX 3060 boasts 3,584 CUDA cores, providing excellent gaming performance for its price range. In contrast, the RTX 3060 Ti is equipped with 4,864 CUDA cores, offering a notable boost in computational power. This translates to superior gaming performance and better support for ray tracing and AI-driven features.

its the cuda

[ornate elk]

yes

more cuda cores

[fair oxide]

so technically 3060ti processing speed is faster

[primal star]

thank you so much for the help!! i did that then got this error?:

[ornate elk]

hey np, how much RAM do you have?

it says memory error

[primal star]

maybe i dont have enough? D:

[ornate elk]

yes thats a bit to less

you would need to increase the windows pagefile to compensate that

or buy additional 8gb ram

[primal star]

ahh thank you!!

[ornate elk]

here is a quick guide:
https://mcci.com/support/guides/how-to-change-the-windows-pagefile-size/

for the windows pagefile increase

make sure to have 20gb free space on C: drive

[ornate elk]

Also i see an AMD GPU there, with that you can use SD without running it on CPU only mode

but you need the AMD Directml Webui Version for that

[runic heart]

has anyone had/seen ongoing issues with SDXL returning blurred out images? Not like a little blurry, but as if a strong gaussian blur has been applied to the entire image.

[onyx solar]

You have some setting wrong somewhere.

[fair oxide]

(comfyui) Is there a way to use the same seed across all generations without a primitive node?

[onyx solar]

Just change it to fixed?

[fair oxide]

I have multiple samplers in my workflow

[onyx solar]

Then using a primitive that is linked to seed converted to input is the correct way to do it.

[fair oxide]

I wish there was a way to connect a primitive node to a reroute node, or just get the seed primitive node to stay the same size on refresh, but I'm not gonna PR it myself 😅

[soft rune]

Anyone know anything about this error? I am using an 1660Ti on Linux, 6gb Vram, 64GB system ram, any other information needed? I checked google and got essentially nowhere, I did a fresh install, still nothing, updated video drivers.

[onyx solar]

pyssss has a reroute primitive node

[short tinsel]

😣😭 I have a problem

I don’t know how to rebring ( dream) to writing the prompts next

[primal star]

thank you!! i think it fixed the memory problem but shows this error now D: not sure if itll work

[harsh token]

Can someone help me with an issue I'm having. So for some reason all of the sudden I can't generate anything because the result is all messed up. Even if I use PNG Info to ensure I'm using the EXACT same settings and prompts as before, the output image is just incoherent mess.

What I think it was is that when I downloaded a VAE and then everything got messed up. But even if I delete it, the problem still persist. I've made sure that all my settings such as clip skip and everything is as it should be. I didn't update or change any driver or Stable Diffusion version or anything like that. This problem literally came up during a session where everything had worked perfectly fine prior minutes ago.

[onyx solar]

Change the VAE setting to Automatic, then change it back or vice-versa.

[harsh token]

I've tried that

[onyx solar]

did you generate after switching, then switch back?

[harsh token]

Yeah

[onyx solar]

So what does the image look like and what are all of your settings?

[harsh token]

This is the image I get. Obviously just incoherent mess. Can't send what it should look like since it's NSFW, but just imagine that the actual image should be coherent at least. Since I'm literally using PNG Info to ensure that.

Not sure if there's a good way to say what my settings are? Or is there anything specific?

[onyx solar]

That doesn't look like a VAE issue.
As far as settings, you need to show a screenshot of all the configuration

[harsh token]

Okay I solved it. Thanks for trying to help, but luckily I got the answer in another server I asked. I had accidentally clicked on Tiling when changing VAE. And I didn't realize it.

[onyx solar]

Yeah, tiling will cause certain output as well.

[harsh token]

Yeah...It was such a minor checkbox that I didn't notice I had accidentally pressed it. I was scratching my head with this issue for an hour lol

[ornate elk]

you should follow my guide fot the AMD install. Its in the pinned messages of this channel

[ornate elk]

Or if you want to run it entirely on the CPU you need to add:
--skip-torch-cuda-test --use-cpu all --no-half --opt-sub-quad-attention
To the webui-user.bat at the line COMMANDLINE_ARGS=

[soft rune]

if anyone finds a solution to this, I don't care if its 2 years from now, feel free to ping me. I've been having this issue for months with everyversion from 1.5 to 1.7 (haven't tried anything lower than 1.5)

[pearl musk]

Yo guys, how higher can a 6600XT go, best I managed at 512x768 was 1.64 it/s on Linux, Arch, Python 3.11.6. But a page claims the 6600XT can do 7 it/s

my args: --listen --no-gradio-queue --port 7862 --enable-insecure-extension-access --medvram --opt-sub-quad-attention --sub-quad-q-chunk-size 256 --sub-quad-kv-chunk-size 256 --sub-quad-chunk-threshold 70

[broken cipher]

Hi

#✍🏼｜rules-and-tos

[oak slate]

could someone please help me😕

[terse sun]

#🤝｜tech-support message

[oak slate]

Nvidia RTX 3070

[unreal valley]

can anyone share a way to change the UI theme of comfy?

[ornate elk]

Thats a connection error.
Make sure you have a good internet speed and stable connection.
If your from Asia you may need a VPN.
Also open up a cmd and type
Pip cache purge
Then delete the venv folder.
And relaunch the webui-user.bat

[ornate ruin]

Hi 🙌🏻 does that look like a clean start to you? I'm trying to find out why my generations take very long on the hires fix part.i have waiting times of over 50 seconds for single images

I7-13700K
64 GB RAM
RTX 4060 Laptop
8 GB VRAM

Sorry! Here is a cleaner boot without any updates making a mess

Windows 11 btw.

[ornate elk]

Hey, set hires steps to 10-15 and enjoy the faster speed

Besides that your Screenshots looking fine

[ornate ruin]

27 seconds at 10 steps
36 seconds at 15 steps
57 seconds at 0 steps

11 seconds without hires fix

It clearly did help thanks! My results right now are very disturbed but that is also at 0 right, will figure that out in a second. Thank you.

Besides that, are those generation timers acceptable?

Realistic Vision V6
DPM++ SDE Karras
29 Steps
Esrgan 4x upscale by 2
Denoising 0.33

512x512 1x1 batches
CFG 4,5
Clipskip 2

Not sure if all of that even effects the average generation time but ..yea 😅😂

[vocal burrow]

steps 0 for hi-res <=> use the same number of steps as the first pass.

[ornate ruin]

So 20 sampling steps means I shall put 20 hires steps.. wouldn't it do just that when left at 0?

[outer trail]

Ok

[ornate elk]

Looks good. You might want to use cfg 7 if your using a normal model

[ornate ruin]

Ah the clue is that you likely would not need the same number of high res fix steps as sampling steps and therefore should put a lower number than you have sampling steps, did I get that right?

[vocal burrow]

"So 20 sampling steps means I shall put 20 hires steps." => no, there's no real point using more than 10/15 steps for hi-res
"wouldn't it do just that when left at 0" => yes like I just mentioned if you let it at 0, in this case it would use 20 hires steps.

[ornate elk]

Yes dont use the same amount of hires steps and steps (so dont set it to 0)

10-15 is mostly enough

[safe igloo]

Where can I make a picture?

[sharp robin]

Will I also need special drivers for my AMD GPU?

[ornate ruin]

Okay good nice that helped. 🙏

[ornate elk]

Nope

[ornate ruin]

Can I also ask here about how my images still look very disturbed right now? Or does it belong in #📝｜prompting-help 😅

[ornate elk]

Sure, mostly if you use denois of 0.7

Change it to 0.5

[ornate ruin]

I was at 0.33 and I raised it to 3.8 now

Can go higher to 0.5 tho sure

[ornate elk]

Depends on the upscaler

Latent upscalers shouldn't be lower than 0.5

[ornate ruin]

Also I changed to DPM++ 2M Karras, felt a bit better

ESRGAN_4x

[ornate elk]

Okay

Feel free to show an example in #📝｜prompting-help

[vocal burrow]

on your computer, on any online website like dreamstudio, clipdrop, etc, on this discord's bots channels when they come back up, etc. Check #1080946152318443610

[oak slate]

I install everything it tells me and it still gives the same error.

[ornate elk]

Follow my install guide in the pinned messages of this channel

[sharp robin]

SDXL turbo isn;t working. I forgot to update my drivers before trying to run it, so that's probably why. I did the --clear_all flag but alas:

Looking into gs://shark_tank/SDXL/mlir/unet_1_77_512_512_fp16_sdxl-turbo.mlir
torch\fx\node.py:272: UserWarning: Trying to prepend a node to itself. This behavior has no effect on the graph.
warnings.warn("Trying to prepend a node to itself. This behavior has no effect on the graph.")
ERROR: Traceback (most recent call last):

also

[ornate elk]

Your using the Shark webui. The best support you get for that is directly on the Shark Discord

[sharp robin]

oh ok thanks!

[ornate elk]

Np, you'll find the link on their github.
If you want to try out the Directml webui fork of auto1111 you can use my install guide in the pinned messages of this channel

[half island]

Hoi, is there a way to make SD/A.I/python shit that eats video memory stop dumping it's nonsense to my pagefile? As every time it starts to do that, i sitll always have at least 15-20GB ram free, but windows/python keeps dumping it into pagefile instead of free ram

[ornate ruin]

Hi:) if I copy the full generation data of an image, is there a way to find out what up scale and denoising has been used? Those seem not to be carried over and are not backed up with most images

Otherwise I don't know how to fully reproduce an image I guess, as the copy data's alone with the right checkpoint and everything does produce a different result.

PNG is not given btw*

[lost crescent]

When did A1111 get to make YAML filesall the time? So time-consuming!!!

[ornate ruin]

I now even did a batch count of 30 with a size of 1 image generation and didn't get the image that I'm looking to just .. copy.. is it not that simple? ..

[warm tusk]

Is it unusual that adding --medvram-sdxl (which I only did to try and be able to go higher when upscaling, that didn't work but I found something that does) when using a 16GB 4060Ti makes no difference in image generation speed? Both with and without that option it takes around 30 seconds to generate a single 1024x1024 SDXL image.

[devout tree]

why all of a sudden i dont have access to bot channels?

[cosmic fog]

Hi

[ornate elk]

it doesnt make yaml files, it just loads them when switching a model or loading one

[unique glacier]

i am currently using comfyui. I got this error while trying to use face detailer.

Error occurred when executing SAMLoader:

`Ran out of input

File "F:\ComfyUI\ComfyUI_windows_portable\ComfyUI\execution.py", line 152, in recursive_execute
output_data, output_ui = get_output_data(obj, input_data_all)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "F:\ComfyUI\ComfyUI_windows_portable\ComfyUI\execution.py", line 82, in get_output_data
return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "F:\ComfyUI\ComfyUI_windows_portable\ComfyUI\execution.py", line 75, in map_node_over_list
results.append(getattr(obj, func)(**slice_dict(input_data_all, i)))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "F:\ComfyUI\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Impact-Pack\modules\impact\impact_pack.py", line 108, in load_model
sam = sam_model_registrymodel_kind
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "F:\ComfyUI\ComfyUI_windows_portable\python_embeded\Lib\site-packages\segment_anything\build_sam.py", line 38, in build_sam_vit_b
return _build_sam(
^^^^^^^^^^^
File "F:\ComfyUI\ComfyUI_windows_portable\python_embeded\Lib\site-packages\segment_anything\build_sam.py", line 105, in _build_sam
state_dict = torch.load(f)
^^^^^^^^^^^^^
File "F:\ComfyUI\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\serialization.py", line 1028, in load
return _legacy_load(opened_file, map_location, pickle_module, **pickle_load_args)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "F:\ComfyUI\ComfyUI_windows_portable\python_embeded\Lib\site-packages\torch\serialization.py", line 1246, in _legacy_load
magic_number = pickle_module.load(f, **pickle_load_args)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^`

[ornate elk]

you need to press the Copy Prompt button on the right corner, there is every information store. paste that into the prompt box and click the blue/white arrow to auto apply

[unique glacier]

Is it okey to ask about comfyui?

[ornate elk]

--medvram-sdxl is made for gpus with 8-11 gb vram, so it doesnt improve speed for your 16gb gpu

[warm tusk]

I was expecting it to slow down the speed. Also don't you need it on a 16gb gpu if using controlnet + xl?

[torn lotus]

hi. how can I install xformers in linux?

[ornate ruin]

Exactly what I did but the result wasn't the same image I saw from where I copied the datas

[ornate elk]

Nope normaly not

But if it takes very long you can try it

[ornate elk]

Using --xformers in the webui-user.sh in the COMMANDLINE_ARGS

[torn lotus]

thanks, I just edited .bat like I was in windows 😄
can I move the stable diffusion directory like it's portable?

[ornate elk]

Yes but that can brake the auto update feature

[torn lotus]

there's no option to correct paths?

there should be an option to tell SD where to find models like CUI

found it --ckpt-dir

[azure magnet]

python 3.10.11 is supported right?

@ornate elk could you help me for a moment?

[ornate elk]

yes

[azure magnet]

why in automatic a1111 say 3.10.6?

[ornate elk]

maybe you installed 3.10.6

[azure magnet]

version: v1.7.0  •  python: 3.10.11  •  torch: 2.0.1+cu118  •  xformers: 0.0.20  •  gradio: 3.41.2  •  checkpoint: fdffd3c312

[ornate elk]

so its 3.10.11

[azure magnet]

I'm freaking out because I'm generating crap with anylora and I don't understand why

[ornate elk]

its not the python xD, can you show an example?

[azure magnet]

the problem it's the aesthetic

[ornate elk]

? looks okay

[azure magnet]

no i mean i'm searching this aesthetic i had tried the checkpoint that suggest me but the output isn't what i want

some use anylora, some use ambientmix

[ornate elk]

okay

maybe try an other upscaler

or the thick outline lora

[azure magnet]

yes, try all of them

[azure magnet]

Is it difficult change python version?

[ornate elk]

you shouldnt change it

it does absolutely nothing in terms of quality or output or speed

[boreal bison]

Hey guys, can someone please help me figure out how to install automatic1111 and Comfy UI?

[ornate elk]

hey, for both check out my Install guides and links in the Pinned Messages of this channel

[fair oxide]

Cones

[boreal bison]

thanks I will now. Which one is better from those 2? Automatic or Comfy?

I need something with full video capabilities

[ornate elk]

if your new to image generating, you better start with automatic1111. much easier to understand and use.
you can also install comfyui, but i would say its for the more advanced users

[boreal bison]

I'm not new, it's been more than 3 years now, and I'm kinda familiar with node based systems, I use houdini for 3d

[boreal bison]

having both on same machine won't be a problem?

sorry for the dumb questions, I just wanna make sure it will work 100% before I start. Had bad experience before

[ornate elk]

nope not at all, you can even set it so that both use the same models, so you dont have to copy models

[boreal bison]

Ok, let me start with Automatic. how much space do I need? Also can I install it on any drive I want? Or it gotta be C:

[ornate elk]

you can install it on any drive, but preferable is an SSD.
Also you need 15-20gb free space on the C drive for cached download files, doesnt matter where you install it. C needs the space only for cached install files.
You can also install the base programm on an drive but later store the large model files on an other drive.

But make sure to install python and git on C

[boreal bison]

Got it!

The very bottom ?

for python

[ornate elk]

yes 64 bit

make sure its 3.10.X (6-11)

[boreal bison]

should I install Git first or Python?

[ornate elk]

doesnt matter

[boreal bison]

It's this, I used the link in the pinned message you have

[ornate elk]

Perfect, my guide covers all steps you need to do, to also get the best performance and compatiblity

for any questions feel free to ask

[boreal bison]

Thank you so much!

I will def need your guidance during the proceess

Python says AMD

my card is NVIDIA

[ornate elk]

thats normal, it doesnt refer on gpu or cpu

[boreal bison]

ahh ok

installing

Does this step have to be C: ?

[ornate elk]

nope, can be any drive

[boreal bison]

awesome!

[ornate elk]

just make sure you make a new folder for it an then do step 2

[boreal bison]

It has to be an empty folder?

[ornate elk]

yes

[boreal bison]

ok

[ornate elk]

whats at the top

[boreal bison]

this

[ornate elk]

copy the command again and paste it with right click

in the cmd

[boreal bison]

ok

seems like it worked doing that

Do I need to update nvidia drivers beforehand?

[ornate elk]

would be a good idea to be updated

[boreal bison]

ok

I'll do it now

does it matter if it's studio or gaming drivers?

[ornate elk]

i dont think that makes a difference

but didnt tested it

[boreal bison]

great

I can't wait to utilize my 3090

After installing this, do I need some custom additional models from other places for video?

[ornate elk]

thats a great card for ai stuff!

[boreal bison]

Good to know! I also have a 3080, can I use them both I wonder?

[ornate elk]

yes, but you cant use them both for one image, but both can run their own instance of auto or comfyui at the same time

[boreal bison]

I don't know who you are, but you're a great guy!

hehe

[ornate elk]

thanks ^^

[ornate elk]

yes, auto1111 and comfyui dont come with video stuff. For that the community made a lot of extensions you can install and try

[boreal bison]

updating my drivers now after which I will proceed with step 5

[dreamy garnet]

stable swarm is also great for multi-gpu setups, it will handle everything and can use comfyUI as backend

[boreal bison]

what is stable swarm? How is it different from those 2 others?

[ornate elk]

its an other webui with a UI similar to auto1111 but uses the backend of comfyui to generate

[boreal bison]

how is it better?

should I install that instead?

[ornate elk]

its still alpha so i wont say its better or recommend it over auto1111

[worthy portal]

Hello to all
I wanted to install the controlnet extension on Stable Diffusion according to the instruction of a YouTube video and I executed the pip install opencv-python command, after that I can no longer open Stable Diffusion and I get this error
.
Thank you for your guidance
.
Already up to date.
venv "C:\AI\stable-diffusion-webui\venv\Scripts\Python.exe"
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Version: v1.7.0
Commit hash: cf2772fab0af5573da775e7437e6acdca424f26e
Launching Web UI with arguments: --xformers --lowvram --no-half
Style database not found: C:\AI\stable-diffusion-webui\styles.csv
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 44, in main
start()
File "C:\AI\stable-diffusion-webui\modules\launch_utils.py", line 460, in start
import webui
File "C:\AI\stable-diffusion-webui\webui.py", line 13, in <module>
initialize.imports()
File "C:\AI\stable-diffusion-webui\modules\initialize.py", line 37, in imports
from modules import processing, gradio_extensons, ui # noqa: F401
File "C:\AI\stable-diffusion-webui\modules\processing.py", line 14, in <module>
import cv2
File "C:\AI\stable-diffusion-webui\venv\lib\site-packages\cv2_init_.py", line 181, in <module>
bootstrap()
File "C:\AI\stable-diffusion-webui\venv\lib\site-packages\cv2_init_.py", line 153, in bootstrap
native_module = importlib.import_module("cv2")
File "C:\Users\Mahdi\AppData\Local\Programs\Python\Python310\lib\importlib_init_.py", line 126, in import_module
return _bootstrap._gcd_import(name[level:], package, level)
ImportError: DLL load failed while importing cv2: %1 is not a valid Win32 application.
Press any key to continue . . .

[boreal bison]

ok so having that won’t give me better results

[ornate elk]

yea it wont

[boreal bison]

cool cool

restarting my computer after drivers update

[ornate elk]

hey, whatever you watched isnt up to date, controlnet doesnt need manually pip installations.

To fix your error try deleting the venv folder first and relaunch the webui-user.bat

[worthy portal]

My venv folder is 5 GB in size, but I will delete it anyway

[dreamy garnet]

will give you the same results as comfyUI is you use that backend, it's a really interesting project maintained by one of Stability's devs

[ornate elk]

thats fine, it will get rebuild when launching the bat again

[worthy portal]

Ok, I am waiting for the downloads and installations to finish

[boreal bison]

Got it

ok resstarted, proceeding to step 5 now

@ornate elk I can't seem to fine this command line in the webui bat file when I edit it

[ornate elk]

webui-user.bat

[boreal bison]

oh snap 😄

2 -- ?

before xformers and after =

and no spacebar?

[ornate elk]

yes

[boreal bison]

ok next step

my 3090 is my secondary card not primary

that's not a problem I hope?

[ornate elk]

then you need to add something to the commandline_args

[boreal bison]

Should I add this?

[ornate elk]

--device-id 1

[boreal bison]

ughh

[ornate elk]

nope

[boreal bison]

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--xformers --no-half-vae

call webui.bat

Can you please modify this? 😦

[ornate elk]

`@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--xformers --no-half-vae --device-id 1

call webui.bat`

[boreal bison]

That's it?

[ornate elk]

128gb ram and i guess a 16+core cpu, whats that beast of a machine xD

[boreal bison]

64 core cpu

64 hyper threaded

I bought it for my vfx work

now using it for fun lol

[ornate elk]

nice, that will run minecraft pretty good i guess 😛

[boreal bison]

I don't understand that game

lol

but I got your joke 😉

[dreamy garnet]

that's a mean machine! houdini sims?

[boreal bison]

yeah, houdini, 3ds max

left 3d world couple years ago though

[dreamy garnet]

i also do vfx, i'm on a 32 core threadripper also 128G ram and 4090

[boreal bison]

trying to create a side business with AI now lol

[worthy portal]

You helped me again, smart boy
Problems solved

[boreal bison]

oh nice!

@dreamy garnet US based?

[dreamy garnet]

no, Europe

[boreal bison]

Just followed you on instagram MetalSupreme

[cobalt quiver]

sighs which version of python do I need to set as my path again, I swear to god every time I do anything else stable diffusion fucks itself

[ornate elk]

3.10.6 up to 3.10.11

[cobalt quiver]

thanks

[dreamy garnet]

pivoted to ai and realtime... mostly doing animation, unreal with comfyUI, some houdini work too

[boreal bison]

@ornate elk stuck here

[boreal bison]

That's awesome!

[ornate elk]

let it run, its doing stuff in the backround

[boreal bison]

I was a lead FX artist in one of big studios here in LA

quit full time like 3 years ago

[cobalt quiver]

I don't see why SD needs the version of python in your path and venv

[dreamy garnet]

awesome

[boreal bison]

Ok! Waiting, fingers crossed haha

[worthy portal]

But something is still left
I can't install controlnet extension
I click on the extension installation, but only loading is displayed and nothing else changes
After closing Stable Diffusion, there is controlnet in the extensions folder, but this time Stable Diffusion will not open.
What is this about?

[boreal bison]

Ok it's done, and the window closed what now? 😦

[ornate elk]

no, thats the actual programm, it needs to be open anytime you want to use automatic1111

[boreal bison]

It said press any key to continue, i pressed enter and it closed

relaunching again

some error I guess?

[ornate elk]

yes

[boreal bison]

No database found?

[ornate elk]

make sure your webui is updated, then check the cmd when installing the extension

[ornate elk]

that can be ignored

[cobalt quiver]

@ornate elk thanks man, I just changed path and it worked 👍

[boreal bison]

What do I do now?

[ornate elk]

you need to delete the rpositories folder.
then go in the stable-diffusion-webui folder. click in the explorer bar (not search)
then type cmd and hit enter
there run
git pull
2 times
then launch the webui-user.bat

[boreal bison]

how to I run git pull from cmd?

just time git pull and enter?

[ornate elk]

yes thats ok

[boreal bison]

runned webui-user again, it's doing it's thing

same error 😦

[ornate elk]

can you make a full screenshot of the cmd?

[ornate elk]

okay then delete the venv folder and relaunch the webui-user.bat

[boreal bison]

on it

[boreal bison]

Some new tab opened in my browser I just noticed it lol

is that it I guess? hahaha

omg I'm generating

[ornate elk]

thats it 🙂

[boreal bison]

Thank you so much!!!

what additional modules do I install now ? haha

Is it using my main 8GB gpu? 😦

[ornate elk]

check the vram usage

when generating

[boreal bison]

that's what I did

3080 is full

3090 is sitting at 10%

[ornate elk]

okay, close the webui, and edit the webui-user.bat again

and set the --device-id to 0