🤝｜tech-support

[rich garden]

is this a 3d or 2d checkpoint?

kk

[jade surge]

pony is for both

[rich garden]

ohhhh ok

had no idea

so you just specify "2d" or "3d" in the prompt?

[ember bough]

GPUs Supported: AMD VEGA, Vii, APUs, RX5600XT, 5700XT, RX 6600, 7700S and above.

[jade surge]

yep

[ember bough]

I guess mine is not supported

[rich garden]

got it

[jade surge]

also you got maybe good gpu ?

[rich garden]

RTX 3060 laptop

i do plan on getting an actual PC soon tho

[jade surge]

i am looking for someone to make me lora

[rich garden]

so is anything with "XL" basically pony?

[karmic crown]

unless you are not on intel mac, in that case welcome to the dark (my) side 🙂

[jade surge]

nope

[rich garden]

oh

[jade surge]

it always says - pony, or something like that

[rich garden]

so what's this then? cuz i don't think i have this checkpoint

[jade surge]

don`t downald this

[rich garden]

oh how come?

[jade surge]

wdm

[rich garden]

u mean don't download?

[jade surge]

ye

[rich garden]

and why not?

[jade surge]

you don`t need

[rich garden]

ok

and what's sd1.5?

[jade surge]

https://civitai.com/models/288584/autismmix-sdxl

[jade surge]

its other checkpoint

[rich garden]

it says pony for base model

[jade surge]

its pony too

[ember bough]

idk what you mean by intel mac im sorry, if you are wondering my cpu it is ryzen 5

[karmic crown]

i had imac with that GPU and intel procesore before

[rich garden]

ok and how can i make it not be pony?

[karmic crown]

macs always use mac stuff even if they have radeon inside

[jade surge]

you can`t

[karmic crown]

thats why i asked

[rich garden]

oh ok

i'm just looking at tryna make some 2d anime style pics like this one

[ember bough]

CS1o is on but dk if he'll help

[karmic crown]

who knows what he is doing, he will answer when he can

[karmic crown]

then download that model and that lora, and use tha prameters from that page

[rich garden]

it uses pony tho right?

[jade surge]

yep

[rich garden]

then yeah i'd have to wait to get a better computer

[jade surge]

you don`t

[rich garden]

oh really?

[jade surge]

i am on 1080 and its working great

[ember bough]

hey cs1o here

[ornate elk]

Follow my AMD Zluda guide. It should work for an rx580

[ember bough]

appreciate it

[rich garden]

it just freezes my laptop every time i try to generate an image and it takes forever to fix

[ornate elk]

Automatic 1111 with Zluda

[ember bough]

can i dm you if i have any problems?

[karmic crown]

you can try with https://civitai.com/models/119012/bluepencil-xl
it is not the same, but it is nice anime model

[ornate elk]

Sure

[ember bough]

ty!!

[jade surge]

maybe its cpu idk

[rich garden]

alright

[rich garden]

maybe

[jade surge]

idk i never made ai on laptop

[karmic crown]

any laptop with 3070+ 8GB should be fine for basic usage

[jade surge]

what about making loras ?

[ornate elk]

@rich garden what was your GPU again?

[rich garden]

RTX 3060

[ornate elk]

But the 6gb mobile one right

[rich garden]

it also takes over 300 seconds for the checkpoint model to load when i select it

mainly the 6GB models

[rich garden]

uhh how can i check?

[ornate elk]

Taskmanager, performance tab, then on GPU, there at dedicated memory

[rich garden]

dedicated gpu memory is 6gb yes

[ornate elk]

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

[ornate elk]

Looks good. And are the models stored on HDD or SSD?

[rich garden]

which is which again?

i think i can check the drives in properties right?

[ornate elk]

Or in the taskmanager

[rich garden]

ok

[ornate elk]

And how much RAM

[rich garden]

where do i check in task manager?

nvm

it looks like they're on HDD

[ornate elk]

Okay, yea that will increase the model loading time

But won't affect the generation speed

[rich garden]

ok

[ornate elk]

And how much RAM?

[rich garden]

on what?

[ornate elk]

System memory

[rich garden]

ah just the normal ram?

16gb

btw what is SD VAE?

[ornate elk]

Ah okay 16gb can be tricky for sdxl or pony.
Then you need to increase the Windows pagefile.

[rich garden]

how can i do that?

[ornate elk]

Copied from aryetis:

launch sysdm.cpl from the Start menu search or run box (Win+R) and navigate to Advanced –> Performances -> Settings –> Advanced –> pagefile/virtual ram -> Modify

https://www.thewindowsclub.com/increase-page-file-size-virtual-memory-windows

[rich garden]

is it safe and even worth doing?

cuz i will be getting a PC later this year hopefully

[ornate elk]

Its safe and worth doing for speed increase when using sdxl/pony models.

[rich garden]

alright

[ornate elk]

You can check the settings. Make sure the pagefile is only enabled for the C drive and disabled for any other drive.
Then adjust the Pagefile Size for C
To Min 16000
And Max 24000
Then save and restart the PC

[rich garden]

does this only apply to a browser?

also i've tried using 2 different 2b loras for yesmix checkpoint and it just comes up with a random anime girl

[ornate elk]

That applys to the PC, so all Programms will benefit from it

[ornate elk]

Hmm recheck if they are compatible with the model.

[rich garden]

so i do that by watching this video? https://youtu.be/UsBfAXiLLAw?si=oaH6oziIOjL0N4YH

[rich garden]

how?

i would assume they're compatible cuz the lora appears in the page when i refresh my lora tab

i feel like if it wasn't compatible then it wouldn't appear

[ornate elk]

Yea but on civitai its always listed what they are made for

[rich garden]

where on the lora page can i check for that?

[ornate elk]

Yes or by following the screenshots

[rich garden]

ok

[ornate elk]

In the info box.
Base model: 1.5 or sdxl

[rich garden]

it says sd1.5

[ornate elk]

Then it should work

[rich garden]

and yesmix seems to be sd1.5 too

[ornate elk]

Maybe try Euler a as sampler and set the lora weight to 0.8

[rich garden]

how do i set it to 0.8?

is that something in the prompt when adding the lora?

[ornate elk]

Yep

Loras in the prompt are
lora:loraname:1

The 1 is the weight

[rich garden]

ok

btw what's the number on the right side of the : ?

like 1:4 for example

the 4

is that like a shortcut for 4 brackets?

[karmic crown]

its lora:loraname:number

it should be something like 0.8 1.4 ... not 1:4

[ornate elk]

Yes (((word))) is pretty much like (word:1.3)

[rich garden]

ok

[karmic crown]

one bracket = 1.1

two brackets 1.1*1.1 = 1.21

and so on...

to be precise 🙂

so 1.331 not 1.3 🤣

[rich garden]

oh and btw not only is the lora and checkpoints etc. on HDD but ALL of stable diffusion is on there cuz it has more storage space

[karmic crown]

i would recommend reorganization if you have enough space

i have 1TB SSD + 1TB micro SD
sd is much faster than HDD, but i still keep everything on SSD

if not, at least put webui and stuff on SSD, and keep models and loras on HDD

[rich garden]

how would the folders have to be if i seperated the models and loras?

[ornate elk]

Then make sure you made the pagefile stuff. And checked that its disabled for the D drive and enables and adjusted for C.

[rich garden]

yeh i just did that

btw the model u linked me didn't work for 2b

[ornate elk]

You can link folder paths to models by editing the webui-user.bat
At the line commandline_args=
you have to add --ckpt-dir "D:\Path\to\myModels"
--lora-dir "D:\Path\to\Loras"

[rich garden]

my stable diffusion folder is 83gb

[ornate elk]

I didnt linked any model

[rich garden]

someone sent this https://civitai.com/models/201351/merge-euler-awe-inspiring

[ornate elk]

Oh I meant Euler A as sampler xD

[rich garden]

i don't get it

[karmic crown]

most of that are models and loras, you can put them on ssd

but put the base files on ssd

[ornate elk]

Your using Auto1111 or Comfyui?

[rich garden]

i only have 151gb free on ssd

[rich garden]

auto1111 i think(?)

[ornate elk]

In txt2img and img2img you can select different sampler. Some work better than other with loras or models

Some are faster and some are slower

[karmic crown]

@ornate elk i do not remember, how much a1111 needs on disk without loras and models?

[ornate elk]

Its above the resolution setting

[ornate elk]

5-7gb

[karmic crown]

i just rememberd, i have test installation (that have 1.5 model inside)

so it requres even less

[rich garden]

didn't get 2b still

[ornate elk]

Do you have an example? How does it look?

[karmic crown]

@rich garden friendly advice: install everything on ssd, but do NOT delete anything from HDD
then just change startup parameters to load loras and modesl from your current install on HDD

[rich garden]

this was with just the lora alone in the prompt

[karmic crown]

you will need just a few gb on ssd

and everyhting will work better

[rich garden]

how do i do that?

[ornate elk]

Follow my install guide for nvidia from the pinned messages of this channel.

[manic flume]

Bit of a simple question probably, is there a way to adjust the frequency of the live preview updating? So like it shows greater or fewer steps as the image generates

whether an extension or a setting

not seeing anything online

[ornate elk]

Thats a setting you can adjust.
Its called show preview every n steps. And at the slider you can adjust that

Should be under preview settings

[manic flume]

this one then?

theres "Live preview display period (in sampling steps - show new live preview image every N sampling steps; -1 = only show after completion of batch)" and "Show the first N batch img2img results in UI (0: disable, -1: show all images. Too many images can cause lag)"

[rich garden]

I clicked the web user ui.bat thing and my laptop froze in the command window (I’m on mobile)

@ornate elk

[manic flume]

guess the first one

[ornate elk]

Yes

[manic flume]

Thx

[ornate elk]

Complete freeze?

Or just the cmd?

[jade surge]

someone can tell me what prompt i have to use to make this guy ?

[rich garden]

my cursor was lagging like crazy and i alreadt restarted my laptop

[ornate elk]

Neymar

[jade surge]

its pagan min

[ornate elk]

I know but he looks like neymar xD

You can always use faceswap to get his face right

[rich garden]

every time i click webui-user.bat after doing ur guide it lags my laptop so i aint doing it

[ornate elk]

How much free space on C ?

[jade surge]

i need hair prompt only

cuz i want to combine him with my profile picture

[rich garden]

160gb

[ornate elk]

Make sure everything else is closed. For example dont run the other webui-user.bat at the same time.

Best is to restart the PC and then launch it

[rich garden]

i already restarted my pc cuz of the freeze

[ornate elk]

Okay please recheck the pagefile settings and screenshot it for me

It shouldn't freeze at all thats so strange

[rich garden]

u mean that C drive thing?

u mean this right?

[quiet gust]

Ok

I’ll show you in a bit cause I’m working out with my mom right now

[ornate elk]

Looks okay

Can you relaunch the webui-user.bat and just let It load until its done. The PC might be lagging hard but if the mouse still can be moved its fine

[rich garden]

I’ll get around to that but it probably won’t be tonight cuz it’s getting late

[lunar bloom]

I'm trying to run SD3 Medium Diffusers in ComfyUI, but I can't get it to work and have no clue why:

`Error occurred when executing DiffusersPipelineLoader:

'time_embed.0.weight'

File "E:\StableDiffusion\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\execution.py", line 151, in recursive_execute
output_data, output_ui = get_output_data(obj, input_data_all)
File "E:\StableDiffusion\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\execution.py", line 81, in get_output_data
return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True)
File "E:\StableDiffusion\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\execution.py", line 74, in map_node_over_list
results.append(getattr(obj, func)(**slice_dict(input_data_all, i)))
File "E:\StableDiffusion\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Diffusers\nodes.py", line 31, in create_pipeline
StableDiffusionPipeline.from_single_file(
File "E:\StableDiffusion\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\lib\site-packages\diffusers\loaders\single_file.py", line 261, in from_single_file
pipe = download_from_original_stable_diffusion_ckpt(
File "E:\StableDiffusion\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\lib\site-packages\diffusers\pipelines\stable_diffusion\convert_from_ckpt.py", line 1436, in download_from_original_stable_diffusion_ckpt
converted_unet_checkpoint = convert_ldm_unet_checkpoint(
File "E:\StableDiffusion\ComfyUI_windows_portable_nvidia_cu118_or_cpu\ComfyUI_windows_portable\python_embeded\lib\site-packages\diffusers\pipelines\stable_diffusion\convert_from_ckpt.py", line 426, in convert_ldm_unet_checkpoint
new_checkpoint["time_embedding.linear_1.weight"] = unet_state_dict["time_embed.0.weight"]`

[quiet gust]

@ornate elk I got it to work but do you know of anyway to speed up generation?

[ornate elk]

What was your gpu again?

[quiet gust]

0.9/7.8

[ornate elk]

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

And make sure to always launch the webui-user.bat and not the webui.bat

[quiet gust]

oh no i can try that

[ornate elk]

You need these for best performance

At the line commandline_args=

[quiet gust]

like are you sure those can boost my performance?

[ornate elk]

Yes

They will absolutely boost it

[quiet gust]

if I add --skip-torch-cuda-test will it affect it?

[ornate elk]

Then it won't work anymore or will be extremely slow

Never add that. Only if your using CPU mode

[quiet gust]

how do I change it?

[ornate elk]

Change what?

[quiet gust]

from cpu to gpu

[ornate elk]

Please Screenshot your taskmanager performance tab

[quiet gust]

cause it says it can't use gpu

@ornate elk you still there?

[ornate elk]

Ahhh

Well you don't have a dedicated gpu

So it uses CPU Mode

[barren rover]

Can someone explain me how sdxl works and if I can have it alongside having sd 1.5?

[ornate elk]

@quiet gust
You need --skip-torch-cuda-test --use-cpu all --no-half --precision full --opt-sub-quad-attention
In your webui-user.bat

[ornate elk]

Its a model like 1.5
So you can download it and put it into the same folder and select it in the webui
But you need to make sure your webui is updated and that you have the right launch args in the webui-user.bat

[quiet gust]

will it speed up the process the same?

[ornate elk]

Should be a bit faster

[barren rover]

how do I know if the webui is updated?

[ornate elk]

Check the version it shows when you launch it

[barren rover]

alr I got the v1.7.0

[ornate elk]

Latest version is 1.9.4

[barren rover]

where can I download the new one

[ornate elk]

You need to git pull manually to Update automatic1111 if you don't have git pull in your webui-user.bat.
Also what's in your webui-user.bat?
Make sure your using xformers
And what's your GPU?

[barren rover]

my gpu is nvidia geforce gtx1650 and heres what I got in the .bat

[barren rover]

Im not being able to git pull for some reason

[ornate elk]

Its using the wrong git from something called nova pasta

In your downloads folder

[barren rover]

Its updated already

I had to drag the git folder into the downloads

now its updated into 1.9.4

[barren rover]

so now I just have to download sdxl and then put it into the models

[sturdy crystal]

Hi guys, I want to try out this colab here for fine-tuning:
https://next.platform.stability.ai/docs/features/fine-tuning

However there seems to be a dependency resolvance issue (see screenshot attached) Does anyone know how to fix it?

[ornate elk]

yep

[barren rover]

can you send me a link to it?

[lunar bloom]

Can someone give me tips on how to properly DL the SD3 Diffusers version and how to make it work in Comfy?

[ornate elk]

here on the right site Filter for any SDXL model:
https://civitai.com/models

[mortal compass]

@ornate elk Evening mate! Do you by any chance use anaconda navigator to switch between python versions?

Btw, thank you for this https://github.com/CS1o/Stable-Diffusion-Info/wiki/Installation-Guides#nvidia-comfyui ! When did you upload all that?

[karmic crown]

i do not recommend anaconda, except for data analysts/scientist
some pacakges might be outdated
and some might be missing
thats the reason i stoped using it few years ago and switched back to pypi

[mortal compass]

Alright, noted! I've seen the name pypi pop up a couple of times

[karmic crown]

install all python versions you need, for example 3.10, 3.11 and 3.12
and then use venv to create virtual enviroments

i have 3.8-3.12 and about 25 venvs 🙈

[mortal compass]

You sir have a problem

[karmic crown]

nope 🙂

[mortal compass]

xd

[karmic crown]

i need to follow requirements

i am using 3.12 for all my new scripts, but sometimes i had to use someting else 🙂

"we can only use 3.7 on that system, change the script" 🙂

[mortal compass]

So when I make all the venvs I need, they will automatically activate when I launch a certain bat file? I'm a bit fresh on the whole programming part

[karmic crown]

if you are talking about a1111, then yes

if you are talking about pyton projects in general, it depends

[mortal compass]

Yeah like the only reason (for now) is that I want to run both A1111 and Kohya

[karmic crown]

what are requiremets for kohya?

3.12?

[mortal compass]

3.11.10

wait lol its 3.10.11

and a1111 is 3.10.6 I believe

[karmic crown]

you can use 3.10.11 for both

[mortal compass]

HOOOOOOOOOL UP

[karmic crown]

i used 3.10, 3.11 and even 3.12 with a1111, but do not try that 🙂

[mortal compass]

Well that makes things much easier

[karmic crown]

unlike some other things, you can have diffrent version of python at the same time on windows, linux and mac

they do not mess eachother

just do not try to have 2 versions of 3.10 or 3.11 🙂

and i would advice to remove conda, unless you are python dev or data analyst 🙂

[mortal compass]

So I take it the easiest way to go from here would be to uninstall anaconda, then get an install exe file of py 3.10.11 and click on upgrade current py installation?

[karmic crown]

i would remove all

install 3.10.11

remove venv from a1111

and rerun webui-user.bat

that will create new one for you

[mortal compass]

Ah right

Anything in appdata I need or want to remove?

[karmic crown]

just uninstall all pythons

with uninstaller

it should be enough

[mortal compass]

Splendid! Thank you for taking time to help me out, fellow viking 😄

[barren rover]

now that I updated sd it is returning this error when I try to generate something

[karmic crown]

what error do you have in cmd?

[chilly trail]

Isn't SD next/Vlad supposed to be like A1 but more cleaned? I haven't used A1 in a long time since I got tired of all the jankyness and weird issues so I moved over to comfyui. Though Comfy is slow to work with compared to A1/Vlad/invoke

[mortal compass]

I think you can run both, no? Or at least have both programs installed and be able to run one program at a time? I'm currently about to install both SD.next and comfy, so I can work with both

@karmic crown got this bad boy after uninstalling all versions of python and anaconda, then installing 3.10.11 and deleting venv

Seems like it's unable to import something that is undefined from pydantic.fields. No idea what that means though

[karmic crown]

i am not sure what your paramteres should be (i am mac user), but try to remove those for now, and leave only xformsers

[mortal compass]

I've run these for about 3 months with no errors, but I can try removing them and remake venv again

[karmic crown]

and please show me this part

################################################################
Launching launch.py...
################################################################
Python 3.12.4 (main, Jun 6 2024, 18:26:44) [Clang 15.0.0 (clang-1500.3.9.4)]
Version: v1.9.4
Commit hash: feee37d75f1b168768014e4634dcb156ee649c05

[mortal compass]

well balls I just closed it lol

I'll show you when I run it again

[karmic crown]

i sad temporary, since i am not sure what they are doing

[mortal compass]

Right, I understand

[karmic crown]

i want to see if a1111 got right pyton version

and if you are on the latest commit

please ignore 3.12, i am testing something

[mortal compass]

Yes of course, I'm following your instructions fully

3.10.11

[karmic crown]

wait, thats forge

ok, that is the last version

btw. i am not sure if you have seen this
https://github.com/lllyasviel/stable-diffusion-webui-forge/discussions/801

they recommend to switch back to a1111 from forge

but it should work anyway

--cuda-malloc (This flag will make things faster but more risky)
--cuda-stream (This flag will make things faster but more risky)
--pin-shared-memory (This flag will make things faster but more risky)

so, yes, as i said, do not use them until you get working enviroment, then try to add one by one

btw. i was made it faster on my mac with "--all-in-fp16 --vae-in-fp32", so you can also experiment with that

[mortal compass]

Sorry, I keep doing rookie mistakes like adding details like "it's forge" lol

[karmic crown]

can you run it without those aditional params? or it failed again?

[mortal compass]

I deleted all the arguments, didn't seem to work

[mortal compass]

Although, I have been meaning to install SD.Next. I'll give it a go with that one

[karmic crown]

try this venv\Scripts\pip install triton

[mortal compass]

That has shown since day 1, apparently it's very common

[karmic crown]

it complains that it miss triton

[ornate elk]

Vlad isnt more cleaned as it tries to support everything and all.
It can be really buggy and not all extensions will work with it

[karmic crown]

@ornate elk do you have any thougts about this?

he removed venv and it just want run

purge cache maybe?

[mortal compass]

installing sd.next..

[karmic crown]

this looks fine for now

[mortal compass]

The error showed up right before launch for forge I believe. It's installing req now, crossing my fingers

[karmic crown]

btw, have you using CS1o's guides?
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Installation-Guides

my advice is to follow his guides for windows

[ornate elk]

Triton is only available on Linux. So this warning is common on windows and can be ignored

[karmic crown]

as i said, listen to CS1o for windows

and if you ever buy mac, just use my script without any guide 🙂

[mortal compass]

Yeah well the problem right now is that we are trying out py31011 for sd.forge and sd.next, and I've always used 3106 with forge

I've not used it before, but you mentioned that 31011 works for both sd and koyha, so I'm trying it out

[karmic crown]

@ornate elk 3.10.11 works fine on windows with forge and a1111, right?

You need to download and install Python 3.10.11 64bit setup.exe from here: https://www.python.org/downloads/release/python-31011/

yep

[mortal compass]

sd.next has booted with no errors lmao

stupid forge

[karmic crown]

try cs1s guide

on the link i sent

[mortal compass]

Oh yeah I've checked out his guides

But the forge one is for the one-click-install file, so that doesn't help me much

[karmic crown]

ok

try this one

got to forge folder and do this
venv\Script\pip install pydantic==1.10.11

that will install pydantic 1.10.11

[ornate elk]

What's the issue?

Forge works with 3.10.11 yes

[karmic crown]

https://github.com/vladmandic/automatic/issues/1670

this one

[ornate elk]

Ah this

[mortal compass]

forge is getting this error when running on py31011

[ornate elk]

put albumentations==1.4.3 and pydantic==1.10.15 in the requirements_versions.txt file
Then relaunch or delete the venv

[mortal compass]

sd.next seems to work fine

[karmic crown]

i told you CS1o knows all about this on windows 🙂

[mortal compass]

AW fuck you've told me that before, I remember now

Was back in april though so excuse my memory xd

[ornate elk]

No problem ^^

[mortal compass]

Thanks!

[mortal compass]

I've bothered him plenty of times, I'm well aware

Forge is working aswell. Thanks guys:)

[mortal compass]

I forgot to ask; so I have sd.next, sd.forge and comfy currently installed. Is it possible to move the root folder from one hard drive to another, or will that make things break? My first thought is to re install on the proper hard drive

[karmic crown]

you can try, but do not forget to change paths to model folders and so on

[mortal compass]

Yeah that is also one of the issues why I need to rearrange stuff, I'll have that in mind

[tame yoke]

is there a way to make comfyui generate the same images as a1111?

[queen marsh]

confused do i need to also download open vino

[scarlet igloo]

Everyone, I have encountered a big problem

[karmic crown]

they should generate same image with same parameteres, sampler, vae, lora...

[tame yoke]

they dont im not sure what im doing wrong
unless a1111 uses a specific scheduler?

[karmic crown]

you need same sampler, scheduler and generator

[scarlet igloo]

I am on the cloud server and it cannot find my model

I've been crazy for hours now, and I still can't solve it. Who knows what I should do

[tame yoke]

its the same, seed too

[brave frigate]

can someone help me?

I'm using controlnet for the first time and for some reason it doesnt effect my generation at all... tho i do have a pose set up.

[brave frigate]

what do i need to put in there? i have nothing to select, i guess i didnt download it

[karmic crown]

you need something like control_sd15_openpose.pth if you are using 1.5

[ornate elk]

Nope only the openvino webui

[karmic crown]

is that sdxl or 1.5 model?

[brave frigate]

sdxl i believe

[karmic crown]

than you need controlnet xl models

CS1o do you have link?

i have it somewhere in backup

[brave frigate]

where can i check which one it is?

[karmic crown]

what model you are using? the base one

[brave frigate]

ah one sec

im using the T-pony, it says base model Pony..

[karmic crown]

huh

i have no idea then if you can use controlnet and how

i only used that with 1.5 and xl

[brave frigate]

maybe it wont work with that checkpoint then

[karmic crown]

i hope someone else knows

[ornate elk]

https://huggingface.co/lllyasviel/sd_control_collection/tree/main

https://civitai.com/models/136070?modelVersionId=267493

[brave frigate]

ah okay so it only works with SD1.5

thx tho that was indeed helpful

[karmic crown]

there are sdxl contronet models

but i am busy now, cant search

[brave frigate]

allgood

[karmic crown]

i was trying to submit new pull request for a1111, and then i encountered some errors

now i am tryng to fix that

[queen marsh]

now its showing me this

[ornate elk]

What did you do?

[mortal compass]

What do you guys prefer, comfy or swarm? I just read on cs1o's guide that swarm is comfy but with a better ui. Might be a better fit for me since I'm inexperienced

[queen marsh]

i just downloaded the open web ui and use the webui batch

[mortal compass]

if you are new to this you need to be very specific when asking for help, like, as specific you can get

[mortal compass]

for example "i went to X website and downloaded X, then I unzipped it in X path, then I clicked on X file, then X happened

[karmic crown]

from which black hole this came?

that version is too old

https://github.com/CS1o/Stable-Diffusion-Info/wiki/Installation-Guides

please follow one of these guides

@queen marsh this was for you

[mortal compass]

@karmic crown do you use comfy and/or swarm?

[queen marsh]

ok ill check it out rn

[karmic crown]

to be honest i am currently not using anything, i have about 10 installations of a1111, since i am trying to make a1111 to work with new torch on mac

i usally use a1111 or forge

but i had comfy installed a few times

[ornate elk]

There is no guide for Intel there

[ornate elk]

Auto1111, but give Swarm a try its like auto1111 ui with comfyui in the background

[karmic crown]

my bad, sorry people

i better shut up, since i am not concentrated to chat

[mortal compass]

I see! Is it still possible to download workflows and such?

[mortal compass]

aight, thanks:)

[ornate elk]

Yep

[mortal compass]

Cool, I'll look into it:) Thanks

[ornate elk]

You have an comfyui tab too with the normal workflow stuff

[mortal compass]

oooh! I'm in

[burnt totem]

hlo i got this from pydantic.fields import FieldInfo, Undefined
ImportError: cannot import name 'Undefined' from 'pydantic.fields' (C:\Users\SYSTECH SERVICING\AppData\Local\Programs\Python\Python310\lib\site-packages\pydantic\fields.py)
Press any key to continue . . . error while installing forge.webui how wi solve this [arc dgpu]

[mortal compass]

I had the same issue today

you need to put albumentations==1.4.3 and pydantic==1.10.15 in the requirements_versions.txt file. Save it, then delete your venv folder. After that, open the webui-user.bat

and it will re-download your venv folder

[burnt totem]

i dont have venv folder in D:\sd\forge.webui\stable-diffusion-webui-forge where i installed it

is it somewhere else?

or i should make one

[mortal compass]

It should be in your root folder. If you don't have one, try adding the things I told you to, then just start the webui-user.bat

It should then start the cmd prompt with something like "creating venv..."

[burnt totem]

got it just wrote the code you mentioned in requirement.txt

[mortal compass]

Please take a scr shot of when you start the cmd, might help

[burnt totem]

the problem is solved thanks to yyou but new one appearedTraceback (most recent call last):
File "D:\sd\forge.webui\stable-diffusion-webui-forge\modules\call_queue.py", line 57, in f
res = list(func(*args, **kwargs))
TypeError: 'NoneType' object is not iterable, it does using xpu

[mortal compass]

Okay, so, you are trying to install forge, no?

[burnt totem]

ya forge.webui. ive 8 gig vramed gpu, i hear it much less vram needable

is the error familiar seems?

[mortal compass]

I only recognise NoneType

Did you follow cs1o's guide for how to install forge?

[barren rover]

@ornate elk now that I updated sd it is returning this error when I try to generate something

also this

[queen marsh]

i still dont know what to do

[burnt totem]

what thats ?

[mortal compass]

I'm not quite sure actually, it might be that the terminal wants a value of something, but the value is not there

Did you follow any guide when installing?

[ornate elk]

hey you need to change --no-half-vae to --no-half

[burnt totem]

its kinda um i mixed, like theres recently support available from openvino for intel gpus to run auto1111 with openvino accelerator scripts, so i tried the first-time-runner.bat file on forge ui and it works like its usin xpu but there no openvino accelerator in scripts so i tried generate image without it and got the error

[mortal compass]

Ooooh I see. Well, I don't know anything about openvino or intel, sorry

[burnt totem]

notto worry man, i asked gpt 4o the bot says i need to change# In call_queue.py

def f(*args, **kwargs):
try:
# Debugging: Print input arguments
print("Args:", args)
print("Kwargs:", kwargs)

    # Call the original function and attempt to convert its result to a list
    res = list(func(*args, **kwargs))
except TypeError as e:
    # Handle the case where NoneType is returned
    if 'NoneType' in str(e):
        print("Function returned NoneType instead of iterable")
        res = []
    else:
        # Re-raise the exception if it's not related to NoneType
        raise
# Return the result
return res............in call_queue.py but where tho

[burnt totem]

hlo i got this TypeError: 'NoneType' object is not iterable how i solve this?

[barren rover]

it worked but now Im also getting this when I try to use my new sdxl model

[ornate elk]

for sdxl models you may need to use --lowvram instead of --medvram, but that also will be slow

[burnt totem]

what bout mine???;[

[ornate elk]

Can't help much with Forge. Its deprecated for normal users

[karmic crown]

I finally figured out whats wrong

https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/16058

now i can make a PR that i intended to post, it works now

[karmic crown]

https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/16059

[trail pumice]

when i am downloading some nodes i encountering some errors
[4:15 AM]
how can i solve exit code 128
[4:15 AM]
anybody knows?

[ruby yoke]

how are you downloading the nodes?

[trail pumice]

from comfyui manager

i tried manually too

[ruby yoke]

can you post the entire error please

[trail pumice]

let me do it again

[crystal mantle]

I spent so long getting past the bugs I had before but now I have this problem, can somebody help me with this?

[ruby yoke]

you at the latest level of git?

[crystal mantle]

pretty sure

[trail pumice]

this is same problem as mine

128 error

[ruby yoke]

yeah. git cant' clone for some reason. what's your syntax?

and how much space do you hve on that drive?

[crystal mantle]

where do I check that

[ruby yoke]

you're running it from manager, so it should be logging the command

[crystal mantle]

the drive I'm trying to install stable on has over 100 GB

[ruby yoke]

can i have the link to the github you're trying to clone?

[crystal mantle]

I don't know It's doing this itself, I've been installing stable since 2 days ago about but kept getting errors and i got to this point

[ruby yoke]

ah. thought you had stable already installed and were installing a node to it. what machine are you installing onto?

[crystal mantle]

Just my pc

everything I've found about this issue just tells you to keep trying but I've been for a while

[ruby yoke]

okay, clean stuff up and delete what you've done, and then go through this tutorial https://youtu.be/Di1KqPXxx2Y?si=4tXjl0ZQNHlDK3vi

[crystal mantle]

or to do it manually but there was no tutorial

great gotta restart

[crystal mantle]

what kinda tutorial is this

[ruby yoke]

it's a step by step how to on installing the SD 3 weights and encoders for comfy. pause at each step and make sure you do eactly what he says

[fair oxide]

Is there a way to use adetailer standalone? For example, if I want to do an img2img workflow that just finds faces and replaces them? But doesn't recreate the entire image?

[crystal mantle]

bro comfy ui looks so goofy

[ruby yoke]

that's a really generic error. let's start by making sure you have enough space on that drive

[trail pumice]

i didnt understood this drive thing

should i check in github destkop?

[ruby yoke]

what machine are you installing to?

[trail pumice]

pc

[ruby yoke]

your PC has hard drives where it writes the files. how many do you have?

[barren rover]

I think my gpu doesnt support sdxl, so I erased it and put on sd again but now im getting this error

[ruby yoke]

that's not much space.

[trail pumice]

why this is not enough?

what is the problem?

[ruby yoke]

your system needs quite a bit of space just to be able to boot your machine on that drive and won't allow you to use it. i'm not sure how large the thing you are trying to clone is, but some of them are fairly large. one thing that can cause the error you're getting is lack of hard drive space. try clearing some space up

[trail pumice]

ok i will try

[dreamy current]

Hi, i just started my local sd-forge and got this message:
extra {'cond_stage_model.clip_l.text_projection', 'cond_stage_model.clip_l.logit_scale'}
left over keys: dict_keys(['alphas_cumprod', 'alphas_cumprod_prev', 'betas', 'embedding_manager.embedder.transformer.text_model.embeddings.position_embedding.weight', 'embedding_manager.embedder.transformer.text_model.embeddings.position_ids', 'embedding_manager.embedder.transformer.text_model.embeddings.token_embedding.weight', 'embedding_manager.embedder.transformer.text_model.encoder.layers.0.layer_norm1.bias',...
How can i fix it

[solemn moon]

for auto sd3, do i need to dwnld the text encoders? wherre do i put them if so

[ruby yoke]

https://youtu.be/Di1KqPXxx2Y?si=zXmM2NXpLy-OMZj4 watch that tutorial, he explains step by step

[solemn moon]

thanks bruh but im asking for Auto1111

[ruby yoke]

i haven't seen a good one for auto1111 😦 - however, yeah, i assume you need to download the encoders. do you have any others already? if so, put them with those

it's not going to hurt to download them.

[solemn moon]

yea already downloaded, just wanted to make sure b4 i destroy things 😅

[ruby yoke]

you put them in the wrong directory, it just won't find them. see if you can find any others already in the system and put them with them. you might also ask on the Banodoco discord

[ornate elk]

SD3 only works in dev branch of auto1111 for now.

[karmic crown]

actually it's sd3 branch, not merged to dev branch yet, but close enough 🙃

[crisp orchid]

Hello, I have a problem. I'm using the "ComfyUI" interface for SD3 and when I try to generate an image after 15 seconds the interface says "reconnecting", when I try to generate it again, there is a "TypeError: Failed to fetch" error. I am new to this business, thank you in advance for your help.

[ruby yoke]

what kind of machine are you running it on?

[crisp orchid]

i5 10400, gtx 1650

[ruby yoke]

you should have a command line window behind the comfyUI browser window. do you?

[crisp orchid]

yea, In the command line, it gives the inscription "pause"

[ruby yoke]

can you copy all the text, from the boot start to yoru error and paste it here

[ornate elk]

Does that GPU has 4gb vram?
Thats to less for SD3

[crisp orchid]

Yes, 4gb

[crisp orchid]

one sec

[ruby yoke]

you only have 4 gig ram?

[crisp orchid]

yes

oh

[ruby yoke]

yeah. you need 16 gig

[crisp orchid]

8gb*

ok thx

gpu has 4gb, system has 8gb

[ruby yoke]

this is mine

[crisp orchid]

https://tenor.com/view/mauro-maurogranate-pov-elmauro-maurogranate14-gif-4545576866420346801

[ruby yoke]

comfy's booting though, but you're not connecting to the server for some reason. you might ask in the #🧣｜comfy-ui channel

[crisp orchid]

ok thanks!

[safe rover]

Following this tutorial: https://www.youtube.com/watch?v=1dnFGxE0vzI
In A1111, when one attempts to produce an animation using the batch input of Controlnet to guide it, AnimateDiff makes an animation without regard for the chosen frames, then removes the motion module, warns that there will then be no temporal consistency, and makes a ton of unrelated images.

The only difference I could find is that his installation lacked a scheduler. Would disabling that fix this?

[ornate elk]

Do you use sdxl motion module?

[safe rover]

I do not.

[ornate elk]

And nope changing scheduler won't fix it

[ornate elk]

Okay then v2 or v3 ?

[safe rover]

v2

V3 needed?

[ornate elk]

Nope

Can you post the full cmd log here?

[safe rover]

One sec

[safe rover]

Wait, I have a script error in there. brb.

[sturdy crystal]

can any1 help me with this?

[safe rover]

To create a public link, set share=True in launch().
Startup time: 23.3s (prepare environment: 5.2s, import torch: 3.9s, import gradio: 1.2s, setup paths: 1.3s, initialize shared: 0.4s, other imports: 1.1s, load scripts: 2.0s, create ui: 8.0s, gradio launch: 0.2s).
2024-06-21 06:37:57,089 - AnimateDiff - INFO - AnimateDiff process start.
2024-06-21 06:37:57,090 - AnimateDiff - INFO - Loading motion module animatediffMotion_v15V2.ckpt from C:\Users------------\stable-diffusion-webui\extensions\sd-webui-animatediff\model\animatediffMotion_v15V2.ckpt
2024-06-21 06:37:58,337 - AnimateDiff - INFO - Guessed animatediffMotion_v15V2.ckpt architecture: MotionModuleType.AnimateDiffV2
2024-06-21 06:38:02,025 - AnimateDiff - INFO - Injecting motion module animatediffMotion_v15V2.ckpt into SD1.5 UNet middle block.
2024-06-21 06:38:02,027 - AnimateDiff - INFO - Injecting motion module animatediffMotion_v15V2.ckpt into SD1.5 UNet input blocks.
2024-06-21 06:38:02,027 - AnimateDiff - INFO - Injecting motion module animatediffMotion_v15V2.ckpt into SD1.5 UNet output blocks.
2024-06-21 06:38:02,028 - AnimateDiff - INFO - Setting DDIM alpha.
2024-06-21 06:38:02,078 - AnimateDiff - INFO - Injection finished.
2024-06-21 06:38:02,083 - AnimateDiff - INFO - AnimateDiff + ControlNet will generate 16 frames.
controlnet batch mode

2024-06-21 06:38:02,371 - ControlNet - INFO - Loading model: control_v11p_sd15_openpose [cab727d4]
2024-06-21 06:38:03,320 - ControlNet - INFO - Loaded state_dict from [C:\Users------------\stable-diffusion-webui\extensions\sd-webui-controlnet-animatediff\models\control_v11p_sd15_openpose.pth]
2024-06-21 06:38:03,323 - ControlNet - INFO - controlnet_default_config
2024-06-21 06:38:05,348 - ControlNet - INFO - ControlNet model control_v11p_sd15_openpose [cab727d4] loaded.
2024-06-21 06:38:05,363 - ControlNet - INFO - Loading preprocessor: openpose_full
2024-06-21 06:38:05,365 - ControlNet - INFO - preprocessor resolution = 512
2024-06-21 06:38:08,169 - ControlNet - INFO - ControlNet Hooked - Time = 6.012529134750366
0%| | 0/20 [00:00<?, ?it/s]2024-06-21 06:38:08,664 - AnimateDiff - INFO - inner model forward hooked
100%|██████████████████████████████████████████████████████████████████████████████████| 20/20 [01:16<00:00, 3.83s/it]
2024-06-21 06:39:28,333 - AnimateDiff - INFO - Restoring DDIM alpha. | 20/440 [01:13<27:15, 3.90s/it]
2024-06-21 06:39:28,333 - AnimateDiff - INFO - Removing motion module from SD1.5 UNet input blocks.
2024-06-21 06:39:28,334 - AnimateDiff - INFO - Removing motion module from SD1.5 UNet output blocks.
2024-06-21 06:39:28,334 - AnimateDiff - INFO - Removing motion module from SD1.5 UNet middle block.
2024-06-21 06:39:28,334 - AnimateDiff - INFO - Removal finished.
2024-06-21 06:39:28,336 - AnimateDiff - INFO - Saving output formats: GIF, PNG
2024-06-21 06:39:30,323 - AnimateDiff - INFO - AnimateDiff process end.
2024-06-21 06:39:30,778 - ControlNet - INFO - Loading model from cache: control_v11p_sd15_openpose [cab727d4]
2024-06-21 06:39:30,780 - ControlNet - INFO - Loading preprocessor: openpose_full

2024-06-21 06:39:30,780 - ControlNet - INFO - preprocessor resolution = 512
2024-06-21 06:39:32,151 - ControlNet - INFO - ControlNet Hooked - Time = 1.5826027393341064
0%| | 0/20 [00:00<?, ?it/s]2024-06-21 06:39:32,255 - AnimateDiff - WARNING - No motion module detected, falling back to the original forward. You are most likely using !Adetailer. !Adetailer post-process your outputs sequentially, and there will NOT be motion module in your UNet, so there might be NO temporal consistency within the inpainted face. Use at your own risk. If you really want to pursue inpainting with AnimateDiff inserted into UNet, use Segment Anything to generate masks for each frame and inpaint them with AnimateDiff + ControlNet. Note that my proposal might be good or bad, do your own research to figure out the best way.
5%|████▏ | 1/20 [00:04<01:31, 4.84s/it]
2024-06-21 06:39:38,309 - AnimateDiff - INFO - Motion module already removed. | 22/440 [01:25<32:40, 4.69s/it]
2024-06-21 06:39:38,309 - AnimateDiff - INFO - Saving output formats: GIF, PNG
2024-06-21 06:39:39,833 - AnimateDiff - INFO - AnimateDiff process end.
Total progress: 5%|███▎ | 22/440 [01:28<27:53, 4.00s/it]
Total progress: 5%|███▎ | 22/440 [01:28<32:40, 4.69s/it]

Sorry for the delay. The lack of a slash between Users and the User name is a weird copy-paste thing. Not in the real logs. The "Total progress 5%" thing ended where it did because I manually interrupted.

[karmic crown]

see this "You are most likely using !Adetailer."

[safe rover]

Manually disabled it.

Do you have to uninstall it?

[karmic crown]

disable should be enough, but you can try to remove temporary and try again

[ornate elk]

Adetailer is not a problem with AnimateDiff normaly

[karmic crown]

why does it have "!" at the beggining? is !Adetailer same thing as adetailer?

[safe rover]

Trying now

Removed Adetailer. Same issue.

*Removed and restarted

[karmic crown]

i do not see any issue in your log, except what i posted

[safe rover]

It keeps unloading the motion module as soon as the first image is animated.

[ornate elk]

!After detailer is the Extension Name in the webui.
Adetailer is the name everyone calls it and the github repo name

Its the same

[karmic crown]

thanks for explaining

but as i said, there are no errors in shown log

[safe rover]

It's supposed to apply the motion module to the batch of images, not try to animate each one.

[ornate elk]

I don't think thats possible with batch input

[safe rover]

That's what the guy in the video does.

[ornate elk]

Oh okay

Make sure your extensions are all updated

[safe rover]

Yup

[ornate elk]

If I have time I will check the video out

[safe rover]

Is there a way to make the program run a check and update all of the scripts and library items?

Deleted venv and had it reinstall everything, but there are a few things that still seem to puzzle SD.

[karmic crown]

you need to have botht at 1

do you have any of those set to more than 1?

[karmic crown]

this is screenshot from the video

[safe rover]

Both are 1

Matched everything except the scheduler, I believe.

[karmic crown]

please send this from yours

actually this one

since you obviously trying this

[safe rover]

Had to manually move things back into position, but I've been going back and forth on these settings since I first encountered the issue.

Hold on. Posting the script errors I mentioned.

ERROR: Exception in ASGI application
Traceback (most recent call last):
File "C:\Users------------\stable-diffusion-webui\venv\lib\site-packages\uvicorn\protocols\websockets\websockets_impl.py", line 244, in run_asgi
result = await self.app(self.scope, self.asgi_receive, self.asgi_send) # type: ignore[func-returns-value]
File "C:\Users------------\stable-diffusion-webui\venv\lib\site-packages\uvicorn\middleware\proxy_headers.py", line 70, in call
return await self.app(scope, receive, send)
File "C:\Users------------\stable-diffusion-webui\venv\lib\site-packages\fastapi\applications.py", line 273, in call
await super().call(scope, receive, send)
File "C:\Users------------\stable-diffusion-webui\venv\lib\site-packages\starlette\applications.py", line 122, in call
await self.middleware_stack(scope, receive, send)
File "C:\Users------------\stable-diffusion-webui\venv\lib\site-packages\starlette\middleware\errors.py", line 149, in call
await self.app(scope, receive, send)
File "C:\Users------------\stable-diffusion-webui\venv\lib\site-packages\starlette\middleware\cors.py", line 76, in call
await self.app(scope, receive, send)
File "C:\Users------------\stable-diffusion-webui\venv\lib\site-packages\starlette\middleware\gzip.py", line 26, in call
await self.app(scope, receive, send)
File "C:\Users------------\stable-diffusion-webui\venv\lib\site-packages\starlette\middleware\exceptions.py", line 79, in call
raise exc
File "C:\Users------------\stable-diffusion-webui\venv\lib\site-packages\starlette\middleware\exceptions.py", line 68, in call
await self.app(scope, receive, sender)
File "C:\Users------------\stable-diffusion-webui\venv\lib\site-packages\fastapi\middleware\asyncexitstack.py", line 21, in call
raise e

File "C:\Users------------\stable-diffusion-webui\venv\lib\site-packages\fastapi\middleware\asyncexitstack.py", line 18, in call
await self.app(scope, receive, send)
File "C:\Users------------\stable-diffusion-webui\venv\lib\site-packages\starlette\routing.py", line 718, in call
await route.handle(scope, receive, send)
File "C:\Users------------\stable-diffusion-webui\venv\lib\site-packages\starlette\routing.py", line 341, in handle
await self.app(scope, receive, send)
File "C:\Users------------\stable-diffusion-webui\venv\lib\site-packages\starlette\routing.py", line 82, in app
await func(session)
File "C:\Users------------\stable-diffusion-webui\venv\lib\site-packages\fastapi\routing.py", line 289, in app
await dependant.call(**values)
File "C:\Users------------\stable-diffusion-webui\venv\lib\site-packages\gradio\routes.py", line 604, in join_queue
session_info = await asyncio.wait_for(
File "C:\Users------------\AppData\Local\Programs\Python\Python310\lib\asyncio\tasks.py", line 445, in wait_for
return fut.result()
File "C:\Users------------\stable-diffusion-webui\venv\lib\site-packages\starlette\websockets.py", line 133, in receive_json
self._raise_on_disconnect(message)
File "C:\Users------------\stable-diffusion-webui\venv\lib\site-packages\starlette\websockets.py", line 105, in _raise_on_disconnect
raise WebSocketDisconnect(message["code"])
starlette.websockets.WebSocketDisconnect: 1001

Not sure how to fix this.

[ornate elk]

Whitelist the webui in any browser adblocker

[trail pumice]

guys this error code: 128 literally happening every node i download

any advertise?

[limber abyss]

to avoid error 128, delete the repositories folder and run the webui. if it still doesn't work, run the webui again and it should work. if it still doesn't, avoid installing to the OneDrive directory and ensure there are no non-ASCII characters in the file path

[sturdy crystal]

Does anyone know how to install stability-sdk in google colab notebook?
running
!pip install stability-sdk
will lead to a dependency problem

[vocal burrow]

what s the full log ?

[trail pumice]

i must delete these?

i am using comfyui btw

[vocal burrow]

dont run comfyui in onedrive folder or any other kind of synched folder

[trail pumice]

where i can export?

[vocal burrow]

export ?

[trail pumice]

I mean, where can I move it?

[vocal burrow]

anywhere else.

c:\ai

d:\tools\ai

e:\MimieMathyTheGoat\ai

etc

[trail pumice]

👍 ok let me try

i will try C disk

[vast fulcrum]

Anyone used OneTrainer for Lora training? How do I resume training? I made sure to save after 10 epoch, but let's say I want to keep training tomorrow, do I just hit start again, but reduce my desired final epoch by 10? Or do I need to add the stopped Lora to the "Lora" tab under "Lora base model". 🐒

[trail pumice]

it didnt work i tried several times

[karmic crown]

use the guide for your system from https://github.com/CS1o/Stable-Diffusion-Info/wiki/Installation-Guides

[random turtle]

Does anyone have success with really accurate IP adapters? I've used Canny Rank 256 and MistoLine both are good but not perfect. Anyone got better results with something else? (I also stacked both of them)

[mortal compass]

If you moved it and it doesn't work, just delete everything and start from scratch.

Any Stable.Swarm enjoyer here? I can't seem to get workflows and nodes (downloaded) to work. Hit me up if you think you can help and I'll explain more

[vocal burrow]

also make sure that you can access github, if you re using some vpn turn it off (unless you re in asia or any "not so pro liberty country")

[pure umbra]

General question : what is the underlying model used for editing tasks ? Is it SD3, SDXL, SD2 ?

[mortal compass]

Can you be more specific? What do you mean by editing tasks

[pure umbra]

inpainting, outpainting

the doc (https://platform.stability.ai/docs/api-reference#tag/Edit/paths/~1v2beta~1stable-image~1edit~1outpaint/post) does not mention which model is in use

[mortal compass]

checkpoints, loras and embeddings I'm pretty sure all play a part in things like inpainting

I know there are specific checkpoints and lora models for inpainting

It's not obligitory to use them though

[pure umbra]

DIffusion models need a special re-training to perform inpainting. Up to recently, the base model was SD2 (a diffusion model). SD3 is a slightly different model (it's not a diffusion but a flow-matching model). I am asking what is the underlying base model when I use SD's API for inpainting tasks.

[mortal compass]

Ooh I see

[pure umbra]

It is not clear from the doc whether inpainting tasks are still done with the SD2 base or the SD3

[mortal compass]

That I do not know I'm afraid

[safe rover]

From bootup to the encountered issue:

Going to reinstall extensions

[safe rover]

No adblocker. UNet seems to be part of this. Could that be it?

[scenic juniper]

hello

[safe rover]

Hello.

[vocal burrow]

might want to update or remove that Perp-Neg-stablediffusion node

[safe rover]

Just saw that. Unsafe?

[opaque sphinx]

is having stable diffusion on M2 disk driver is diff from having it on normal ssd or even hdd ? does it affect any process ?

[vocal burrow]

what do you mean unsafe ?

[safe rover]

Was asking

[vocal burrow]

it's throwing an error so yeah kinda, ignoring an error is not good practice

[vocal burrow]

The slower the device holding your model is, the slower it will transfer the model to the gpu.
Outputs will not differ, nor will it affect the generation speed once the model is loaded on your gpu (assuming you're not using --lowvram --medvram or similar stuff)

[safe rover]

This cut out all of the startup scripting errors. Didn't change the main issue, but I'm beginning to believe the guy in the vid is just leaving out something important. Moving on from this.

[unique wasp]

Greetings, I am not sure if this is the right place to ask this question, but I recently installed the "TensorRT" extension within A1111 and did the "export engine" for a model, but it now takes 6x as long generating an image.

[cursive cobalt]

I'm BAAACK! Ignore the deformaties.

My Comfy workflow isn't incorperating my lora's, after a fresh install (Yes, i did "Install missing").

[burnt totem]

getting this Prompt executed in 300.27 seconds timeout error in comfy. while like downloading wd-vit-tagger-v3.onnx model @ornate elk is it set the timer somewhere? if so how i change that?

[soft fiber]

what happens when you only use 1 lora?

[cursive cobalt]

So. I went and made a brand new flow. 1 ckpt, 1 vae, 2 lora's.
Everything works fine on that flow, so it seems. Something is corrupted in my old flow

[barren rover]

@ornate elk I keep getting this error now

[burnt totem]

you should lower the loads eg. resolution the memory yuo have isnt enough

[cursive cobalt]

Do smöl resolution then upscale :3

[barren rover]

what are the loads eg?

[burnt totem]

eg. example

[barren rover]

oh ok, so what are the loads that I have to lower

[burnt totem]

resolution

[unique wasp]

Just in case here's a screenshot of my TensorRT problem.

[ornate elk]

How much RAM do you have?

You may need to increase the windows pagefile

[barren rover]

I have 8 gb ram

[barren rover]

how do I do that?

[cursive cobalt]

8? yikes

[burnt totem]

getting this Prompt executed in 300.27 seconds timeout error in comfy. while like downloading wd-vit-tagger-v3.onnx model is it set the timer somewhere? if so how i change that?

[ornate elk]

Copied from aryetis:

launch sysdm.cpl from the Start menu search or run box (Win+R) and navigate to Advanced –> Performances -> Settings –> Advanced –> pagefile/virtual ram -> Modify

https://www.thewindowsclub.com/increase-page-file-size-virtual-memory-windows

[unique wasp]

TensorRT may not be very popular in general,where could I ask my question, are there discord channels that are more in-depth regarding this extension?

[junior trellis]

What modell does midjourney uses

🙂

[trail pumice]

guys is this alright

when i download smt from github

download speed slowing

[noble belfry]

depends on your plan. ive a stupid priced broadband plan that never reaches the speeds its supposed to and im not even miles from the nearest exchange. bigger use at weekends means some networks just hobble you. aria2 might help, its a command line downloader that tends to optimize downloads

ive never got blazing fast speeds from github anyway, but 9/10 projects are just code and im not downloading full binaries

[trail pumice]

nice explanation

thank you

[terse urchin]

#🤝｜tech-support Can anyone help I had some problems with stable diffusion running slow so i deleted and reinstalled the venv file and now when i try and use ip-adapter-faceid it wont work and says RuntimeError: C:\a_work\1\s\onnxruntime\python\onnxruntime_pybind_state.cc:866 onnxruntime::python::CreateExecutionProviderInstance CUDA_PATH is set but CUDA wasnt able to be loaded. Please install the correct version of CUDA andcuDNN as mentioned in the GPU requirements page (https://onnxruntime.ai/docs/execution-providers/CUDA-ExecutionProvider.html#requirements), make sure they're in the PATH, and that your GPU is supported. i know I'm missing something obvious

[noble belfry]

sounds like a version/pathing issue. hooray. is SD itself running ok?

[ornate elk]

hey, whats your GPU?

[sullen hound]

Hey all, i am hoping this is ok to ask here. I am trying to use ComfUI + ComfyUI-DynamiCrafterWrapper to create an animation. I do not have a great system. 32gb RAM and 8gb VRAM on an nvidia 1080. I am able to run and get the animation, but it always looks like its melting. It reminds me of a vae issue. I'm not sure what i am missing. It appears to be giving me a corrupted frame for each frame i tell it to give, including the three frames that i want but they look like trash. I am resizing the image to 248*248 because that was the highest i could get it to run at. Ive tried all different frame counts and frame rates, loop count etc with the same result. I might be getting those wrong too as i likely am not balancing them correctly.

[sullen hound]

this is the best output video i can get after changing the input image sizes to 256x192 so maybe my issue is image size and then ultimately my vram? idk anymore

[terse urchin]

Yeah no problems otherwise

[terse urchin]

Rtx 3090 24GB

[ornate elk]

whats your webui version?

[terse urchin]

I should have left things as they were but thought a reinstall would speed things up, which it did but now this 😤

[terse urchin]

Where do i find that?

[noble belfry]

should say at the bottom of the screen when you load it up, as a footer. at least for a1111

[tame yoke]

how do i update forge to the version before the announcment

[noble belfry]

yeah honestly, it cant do anything with that size of image. even the older sd1.5 models are trained on 512x512 anything too small is going to come out garbled, not just low resolution. its not a perfect analogy, but think of it as trying to stamp in big squares of what its trained on. if its too small its going to be a very muddled mess.

[sullen hound]

dang ok. I kinda thought i was moving in that direction when i increased it and it looked 10x better

[noble belfry]

https://github.com/lllyasviel/stable-diffusion-webui-forge/releases should list older releases for you

[sullen hound]

i appreciate the answer

[noble belfry]

8gb vram should let you do something though. might melt your pc but still 😄

[sullen hound]

ill let it melt if it means i can get something done, i mean i can always just generate the frames by hand and combine them in an editor the old fashioned way

[tame yoke]

it only shows latest + i tried clicking on the release before the announcment and ‚download as zip’ but i think it downloded the source code

[noble belfry]

well its written in gradio/python so its going to be source code 🙂

[tame yoke]

without the run bat and stuff

it spit an error

and downloading from releases worked

but i had to update so

[noble belfry]

you want this one most likely

just as an aside, its probably better to move back to a1111 for the future, or at least have both

[broken tulip]

I want to use a node that does a lens blur, however it has only one image input. How would I use the ZDepth to control it? Is there a way to use the zdepth as a mask?

In a composite or something?

[noble belfry]

probably better asking in #🧣｜comfy-ui for node lego stuff 🙂

[broken tulip]

Oke thanks 😛

[noble belfry]

i mean, someone in here might know, but they'd know faster mate 😄

[broken tulip]

I know how to do it in PS and Nuke, just wantg to have a Comfy workflow for faster testing

[tame yoke]

the one i used but, updat.bat moves me to the newest and if dont i have a really old one

and i use comfyui, i just wanna see how fast i can get since i really dont use fancy stuff so im fine using deprecated stuff

[pearl musk]

Hey people, I'm trying to learn how to use Regional Prompter, I've watched at least 4 different tutorials and none helped me. I want to use Regional Prompter to generate twins, they have the same hair and eyes, but they wear different clothing. I have tried everything with RP, and yet I can't make it work (obviously skill issue lmao). I only get one girl/boy with mixed clothing. I am on sd-webui-forge, using SDXL models with booru prompts.

[dusky blaze]

Should join the "Banodoco" server - it's all about AnimateDiff and other SD animation stuff. Kijai who coded the wrapper you menitoned chats in that server for ex.....

[burnt totem]

does controlnet sd 1.5 work with sdxl checkpoint

if now how much vram does it needs to sdxl controlnet to works with sdxl checkpoint

[ornate elk]

At the bottom of the webui

[gaunt goblet]

Hey, sorry to bother but recently i've been trying to get stable diffusion forge working for myself but I have AMD driver, I've tried installing the HIP SDK for windows 5.7 but it fails... is there any sort of solution? the only error im getting is 215 and when I go try and fix it, it doesnt seem to have any information on it.. and I'm not THAT good with file digging to find it

if anyone would have any solutions or be willing to talk to me about it I would appricate it

[gaunt goblet]

Traceback (most recent call last):
File "E:\Forge AMD #2\stable-diffusion-webui-amdgpu-forge\launch.py", line 51, in <module>
main()
File "E:\Forge AMD #2\stable-diffusion-webui-amdgpu-forge\launch.py", line 39, in main
prepare_environment()
File "E:\Forge AMD #2\stable-diffusion-webui-amdgpu-forge\modules\launch_utils.py", line 603, in prepare_environment
from modules import devices
File "E:\Forge AMD #2\stable-diffusion-webui-amdgpu-forge\modules\devices.py", line 3, in <module>
import ldm_patched.modules.model_management as model_management
File "E:\Forge AMD #2\stable-diffusion-webui-amdgpu-forge\ldm_patched\modules\model_management.py", line 122, in <module>
total_vram = get_total_memory(get_torch_device()) / (1024 * 1024)
File "E:\Forge AMD #2\stable-diffusion-webui-amdgpu-forge\ldm_patched\modules\model_management.py", line 91, in get_torch_device
return torch.device(torch.cuda.current_device())
File "E:\Forge AMD #2\stable-diffusion-webui-amdgpu-forge\venv\lib\site-packages\torch\cuda_init_.py", line 778, in current_device
lazy_init()
File "E:\Forge AMD #2\stable-diffusion-webui-amdgpu-forge\venv\lib\site-packages\torch\cuda_init.py", line 293, in _lazy_init
torch._C._cuda_init()
RuntimeError: Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx
Press any key to continue . . .

[burnt totem]

its mee mario,........ ive been trying to install forrge too but with intekl arc gpu unsurprisingly i got lots of error

[burnt totem]

use this --directml in the argument maybe work?

[gaunt goblet]

Where would I have to put it? In the user bat?

[tall shard]

Can I run this server without gpu ?

[burnt totem]

yh go to edit the webui.user.bat the next to argument also first delete the venv, im not sure if it work

RuntimeError: mat1 and mat2 must have the same dtype, but got Float and Half
mat1 and mat2 must have the same dtype, but got Float and Half
*** Error completing request
*** Arguments:
Traceback (most recent call last):
File "D:\sd\forge.webui\stable-diffusion-webui-forge\modules\call_queue.py", line 57, in f
res = list(func(*args, **kwargs))
TypeError: 'NoneType' object is not iterable

--- im gettinng this @ornate elk how do i solve it

[gaunt goblet]

does this go in command ARGS?

[burnt totem]

the same way as --skip-cuda

[gaunt goblet]

got it

Trying it

[gaunt goblet]

but I was curious,, I got it working for normal but stable forge did not work

new errpr

[burnt totem]

the forge bit complicated then auto11 i also somehow run auto with intel but not forge

[gaunt goblet]

what ideas do you suggest? if one

[gaunt goblet]

Im going to head to bed, if anyone can respond to this later please, been at a loss at it for a bit

[sleek escarp]

I'm new to comfyUI and am trying to get a node to work, "Load IPAdapter (SDXL Plus)", but every time i try to queue a prompt, i get the error "Prompt outputs failed validation
IPAdapter: - Required input is missing: ipadapter"

the problem is there is no way i can see to input an ipadapter

this is the workflow i am trying to use (drop the image into the UI)

idk if discord messes with the metadata needed to make that function work

its from here: https://github.com/nerdyrodent/AVeryComfyNerd?tab=readme-ov-file

[thorny dragon]

@ornate elk hey there bro, its' me again 🙂 I was trying out the regional prompter extension in a1111 and it gives me a "list index out of memory" error. Why's this happening?

my bad....it says "list index out of range".... not "list index out of memory"

[ornate elk]

Don't use forge. Follow any other AMD install guide from my guides in the pinned messages of this channel.
What's your GPU?

[ornate elk]

Hey, can you try without the mask mode but with matrix?

[ornate elk]

Mostly that error is caused when using a model and lora that are not compatible

You need to also use --no-half

[thorny dragon]

yeah matrix mode works, but i want to use mask mode for more precision...i am not using any lora and this model works with regional prompter too...

[thorny dragon]

oky, let me do that and tell u the result....

[ornate elk]

No that wasn't for you

[thorny dragon]

oh sorry...didnt see that 😅

then what should i do? 🙃

[ornate elk]

Idk, I would need to test mask mode for myself

[jade surge]

is there any lora that makes only like smile reference ?

[thorny dragon]

plz do bro, would be a great help.... a1111 seems so buggy and slow recently...any updates on when version 2.0 is coming?...i heard its supposed to be as fast as forge...

[ornate elk]

Idk when its coming, but I guess maybe the next weeks

[trail pumice]

guys anybody knows how to fix it, i literally tried everything

i downloaded ffmpeg to this path

but it is still giving me error

[untold urchin]

Hi everyone,
[checked chat history and decided to post my problem here]
Yesteday I was installing an upscaler extension and then after restarting the WebUI, SD stopped working. I click on the bat file to start SD again, it gets stuck in "Commit hash: feee37d75f1b168768014e4634dcb156ee649c05".
When I Ctrl+C I see the error below:
"File "C:\Users\yuva\stable-diffusion-webui\modules\launch_utils.py", line 236, in run_extension_installer
stdout = run(f'"{python}" "{path_installer}"', errdesc=f"Error running install.py for extension {extension_dir}", custom_env=env).strip()"

Sharing a full screenshot to make sure I am not missing anything.

I realized that it works [this morning] if I run it without internet connection.

I deleted:

• The cache folder
• The problematic extension file

Not fixed my problem.

Any help would be much appreciated

[ornate elk]

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

[untold urchin]

will try thanks. But I see that the folder size is big and I only have my 4G internet ATM :/ So I can't redownload the folder I am afraid.

[ornate elk]

ohh okay, then you should do that when you have ethernet of wifi

[ornate elk]

i tested the function, and its very complicated. it worksd with one mask but not with two

[untold urchin]

It works now after deleting the extensions folder. I did a backup though and now will try to install them 1 by 1

[ornate elk]

whats the extension you tried to install?

[untold urchin]

That's the interesting part of my story.
I tried to install this yesterday https://github.com/pkuliyi2015/multidiffusion-upscaler-for-automatic1111
Then the problem occurred. Now even though it is deleted, my WebUI still did not start. As if it broke something yesterday. Although I fixed my problem now, I don't know what is the source issue and whether it will repeat again today

[ornate elk]

ah that extension should work normaly, i have it too

[untold urchin]

then a coincidence happened and I will cry again in the near future 😄

and these are the extensions I have [now deleted for safety]

[ornate elk]

yea, best is to delete the venv and then install the extension again

[ornate elk]

they are all fine. shouldnt make any problems (only maybe the prompt generator) i dont have it

[untold urchin]

alright. do you recommend adding auto git update param to the launcher?

[ornate elk]

i have git pull in the webui-user.bat
and i also have --always-update-all-extensions
in the launch args.
But these can also break your webui when a broken updates gets pushed. Its up to you.

[untold urchin]

hmm alrighty. Tnx for the help! Will cry here if the problem persists , hopefully not.
Here is a cookie for the help
🍪

[ornate elk]

no problem 🙂 yea let me know if it worked

[untold urchin]

found it @ornate elk !
sd-webui-controlnet causes the issue.
Everytime I delete controlnet extension folder SD runs normal. Whenever I copy it back, it gets stuck and doesn't launch the WebUI. What is interesting is that I have been using these extensions for months. It only happened yesterday. I believe an update broke it.

So does this mean the venv folder is "toxic" ?

[ornate elk]

best is to launch the webui without controlnet extension and then reinstall it via the extensions tab.
Controlnet doesnt make an issue for me right now. launched without problems

Extensions can mess up the venv folder ans they wont fix it when beeing deletet.

so to be "clean" you need to delete the venv folder at least once after deleting extensions to be sure

[terse urchin]

i have this number Commit hash: feee37d75f1b168768014e4634dcb156ee649c05 and this info version: v1.9.4  •  python: 3.10.6  •  torch: 2.1.2+cu121  •  xformers: 0.0.23.post1  •  gradio: 3.41.2  •  checkpoint: 24a393500f

[ornate elk]

thats the latest version 👍

[terse urchin]

i have a copy of what comes up when i try to use it aswel but its long

[ornate elk]

sure paste it in here

[ornate elk]

okay first check for extension updates

in the webui, in the extensions tab

then apply them and delete the venv folder

because dynmic prompts seems to have an issue too

Then make sure you have the latest nvidia driver installed for your GPU

[terse urchin]

After this issue started i updated nvidia drivers yesterday so thats one thing on the checklist I'll do the others now

[junior trellis]

cant find ip adapter here

[ornate elk]

you need to download the ip-adapter models from here:
1.5 ones: https://huggingface.co/h94/IP-Adapter/tree/main/models
SDXL ones: https://huggingface.co/h94/IP-Adapter/tree/main/sdxl_models
Face-ID: https://huggingface.co/h94/IP-Adapter-FaceID/tree/main

[junior trellis]

but i have

[ornate elk]

click on IP-Adapter

also make sure the extension is updated

[junior trellis]

something strange also why did my canny and others dissapear from the models folder

?

i added all these but now they are not here

[ornate elk]

maybe they are in extensions/sd-webui/controlnet/models

[junior trellis]

ah thanks did not know files can be moved by themselves

[terse urchin]

Ok so extentions are now updated and the venv is deleted and currently reinstalling

[night spoke]

Hello, anyone knows how to solve this problem?

I installed the SD step by step with the tutorial video, but still can run it normally

[ornate elk]

hey, whats your gpu?

[night spoke]

4060ti

Nvidia

[ornate elk]

then you need to edit the webui-user.bat
at the line Commandline_ARGS=
you need to add: --xformers --no-half-vae
then relauch the webui-user.bat

Also make sure the webui is whitelisted in any browser adblocker

[night spoke]

like this?

[ornate elk]

yep

[night spoke]

an d how can I check if its whitelisted?

[ornate elk]

what adblockder do you use?

[night spoke]

no adblocker

I clicked the webui-user.bat again

it's been a while. Should I wait for installing?

[ornate elk]

yea just let it load

[night spoke]

Okay, thank you so much. You're awesome!

[manic flume]

Do people normally use upscaling with Pony models? Seems to cause my generation to stop at 50% using ERSGAN_4x havent tried the other options

[ornate elk]

yes upscaling works fine with them

[ornate elk]

how much hires steps?

[manic flume]

0

Im trying with Latent

Latent seems to stop at 50% too

Wait nvm

Still going

Slowly but its going

Might just be a lot for the gpu and perhaps ERSGAN_4x is just too powerful to handle, idk

Using a rtx 3060

Oh i see

Its upscaling by 2 so going to 2048x2560

Thats quite a bit

[night spoke]

Em...same prompt, what should I do next?@ornate elk

[ornate elk]

what settings do you use? and which browser?

[ornate elk]

always set them to 10, if not it will take really long and can crash

[night spoke]

Chrome

set up what to 10?

[manic flume]

Okay

[manic flume]

He's talking to me about the hires steps

[night spoke]

Ok

[ornate elk]

okay and can you try an other model?

[terse urchin]

error still happening

[night spoke]

No, I can't. The checkpoint says processing

[ornate elk]

how much RAM do you have?

[night spoke]

32GB

[terse urchin]

this is everything shown

[ornate elk]

you can also speedup sdxl/pony upscaling by using the Tiled VAE extension and using the sdxl fp16 vae

[ornate elk]

thats good. which install guide did you followed?

[terse urchin]

other ip adapters work just not faceid models

[manic flume]

Thanks for the tips

[night spoke]

@ornate elkThis is the video link: https://www.youtube.com/watch?v=onmqbI5XPH8

[terse urchin]

could it be my CUDA?

[ornate elk]

you mean they dont show up in the drop down?

face id models need also the face id loras

[terse urchin]

they error out

i have the loras

[ornate elk]

okay

maybe its an controlnet issue, because i testedt it rn and got the same error with face id.

[terse urchin]

oh i thought it was me because i first got it after a reinstailed venv yesterday

maybe i need to roll back auto1111

[ornate elk]

maybe it can be fixed by uninstalling onnx and installing the nvidia version:
close everything then
go into the stable-diffusion-webui folder
click into the adress bar and type cmd and hit enter.
then enter these three commands one by one:
venv\Scripts\activate.bat pip uninstall -y onnxruntime onnxruntime-gpu pip install onnxruntime-gpu

[ornate elk]

okay then please try an other browser

like Firefox or Edge

[night spoke]

How to set up other brower for launching, it turns to chrome in default

[ornate elk]

copy the url and paste it into the other browsers adress bar

it opens only in your primary browser by default

[night spoke]

@ornate elkOkay, I just pasted the url on Edge. It has the same issue in Chrome.

[ornate elk]

can you share the txt2img settings you use? like prompt, model, resolution etc?

[night spoke]

@ornate elk

[ornate elk]

okay, close the webui and cmd.
go into stabel-diffusion-webui/models/stable-diffusion
there move all the models into an other temporary folder.
then download this Dreamshaper model:
https://civitai.com/models/4384/dreamshaper
and put it into that folder.
then relaunch the webui-user.bat

[night spoke]

@ornate elkokay, I'll try

[terse urchin]

did that, same error still occuring

well looks the same at least

[honest tendon]

this is so sad

no gpu

https://civitai.com/articles/4170/comfyui-vid2vid-all-in-one

will this work on AMD?

6950xt

[terse urchin]

Thanks for helping, I'm unsure what else to do

[ornate elk]

if your using a realistic model you can use Reactor extension for faceswap instead

[terse urchin]

Reactor is one i use, i just like the style of ip-adapter-faceid. Can't figure out what's wrong. Error mentioned PATH but everything is where it should be

[ornate elk]

maybe open up an issue on the controlnet github. including screenshots and a description
https://github.com/Mikubill/sd-webui-controlnet

[terse urchin]

Ok thanks, is the error still happening for you aswell?

[night spoke]

@ornate elkHey, I downloaed the new model, but the issue still exists

[ornate elk]

does it also loads endless in the model selcetion dropdown?

[night spoke]

@ornate elkyes

[ornate elk]

do you run any third party antivirus ? or any firewall browser extension?

[ornate elk]

yes, but im on AMD, still getting an error only with face-id related to onnx

[night spoke]

@ornate elkonly Windows Defender

[ornate elk]

okay, try deleting the venv folder and relaunch the webui-user.bat

[night spoke]

@ornate elkstill loads forever

[finite zenith]

Do you really need to crop images to 1:1 for training loras?

[burnt totem]

got this, from pydantic.fields import FieldInfo, Undefined
ImportError: cannot import name 'Undefined' from 'pydantic.fields' (D:\sd\forge\stable-diffusion-webui-forge-main\venv\lib\site-packages\pydantic\fields.py)
Press any key to continue . . .how do i solve this

[ornate elk]

Okay then please reinstall the webui using my guide from the pinned messages of this channel

[ruby yoke]

i would.

[lost crescent]

In ComfyUI Queue Prompt keeps disappearing?!?!?

[real goblet]

Hi people, I got a quesion, all of my outputs come out looking very questionable (discolored, etc) What might be causing this?

[ornate elk]

Hey, Mostly that's caused by using a Wrong VAE or no VAE at all

[real goblet]

So hold on just so you can get a look at it I'll make a quick gen

[burnt totem]

heloo @ornate elkso ive successfully launched and generated image with foirge.webui with xpu ipex, but the vram usage is 7.9 gb this isnt what the creator promised, any idea how i optimize this?

[real goblet]

Alright so this is the image I tried recreating, used the same settings and promp and model and all and below is my output to compare to. https://civitai.com/images/6407032

And I only get really bad gens on Windows, on Linux it works like a charm

but I am lazy and I don't really want to reboot into the other OS

[ornate elk]

What settings did you used ?
Also what's your GPU and what's inside your webui-user.bat?

[ornate elk]

You can checkout the forge github for optimisation commands. But I don't think any cuda optimisations will work with Intel

[real goblet]

RTX 3070, the non-locked one

[burnt totem]

i tried but it automatically removed any command i use and says "it is for optimization"

[ornate elk]

The webui-user.bat not the webui.bat

[ornate elk]

Okay thats strange

But forge won't get fixes or updated anyway. Better try getting openvino or SDnext to work

[ornate elk]

Oh your using SD Next?

[burnt totem]

talk about sdnext the sdnext-modernui
wont enabled like i tick mark and apply changes its reboot then the tick is unchecked and the ui is that old one

[real goblet]

Yeah, normal SD gives equally bad results so I tried both

[ornate elk]

Okay please show me the webui-user.bat of auto1111.

[real goblet]

Its just the default one

[burnt totem]

what bout mine error man

[ornate elk]

Okay at the line commandline_args=
You need to add:
--xformers --medvram-sdxl --no-half-vae
Thats for the best performance and compatibility.

Then save and relaunch the webui-user.bat

[ornate elk]

You need python 3.10.11 64 bit

SD webui doesn't support 3.9 or 3.11/3.12

[real goblet]

Interesting, since it worked prior

Is it possible to force a specific python ver?

[ornate elk]

Yea it can work but some features or extensions wont

[ornate elk]

If you have multiple versions installed yes.

Edit the webui-user.bat at the line
PYTHON=

there add the path to the python.exe of 3.10.11

[real goblet]

Yeah, since I got quite a lot 😅

[ornate elk]

After that you need to delete the venv folder

[real goblet]

Alright

[real goblet]

It still complains, despite I've specified python ver

[burnt totem]

hlo

[ornate elk]

Recheck the webui-user.bat

[real goblet]

Well, it seems about right to me

[ornate elk]

Hmm yea but idk if that works with that conda path. It says Vlad and is already used in a venv

So maybe that won't work

Better to install python 3.10.11 the normal way and add that to path

[real goblet]

I'll try that then

[ornate elk]

Then delete the venv folder again

[real goblet]

Making a gen, but the reason why I would want to use vlad over normal sd is because its mega fast

normal sd takes multiple minutes just for one gen

[ornate elk]

That should be fixed now, you missed the launch arguments

Now you have them --xformers --medvram-sdxl

[real goblet]

Well, I do have them

Its that slow despite the args

[ornate elk]

How slow?

[real goblet]

Over 5 mins just for one image

[ornate elk]

Can you show me the settings you try?

Like resolution, steps model etc

You can screenshot it

[real goblet]

Really fucked result as well

[ornate elk]

Set the normal steps to 30
Set the denois to 0.5
Set the hires steps to 10
set the upscaler to ESRGAN 4x+
Set the scheduler to automatic

Then it should be fine

[real goblet]

So its mega fast until 48% then it basically freezes

[ornate elk]

Thats the hires steps

"Upscaling"

It shouldn't freeze

Make sure every other programms are closed like games and Wallpaper engine

[real goblet]

Yeah that it a lot better already, but bro truly got man boobs 💀

[ornate elk]

Oh xD

So now to your initial issue "how to get the same image like on civitai"

@real goblet
To get the same image like on civitai is not alwaye possible but make sure you do it like this:
On the image in civitai click on the blue copy prompt Settings button.
Then paste that into the positive prompt.
After that Click on the blue/white arrow at the right.
It should now apply everything automaticly.
Then generate.

[ember bough]

Is there a limit for Automatic1111 as model bases?

Can I only used SD 1.5 based ones?