🤝｜tech-support

[hot thorn]

its all SD and how to min max this xD

thoughts are linear af xD

btw, it seems that mollac was faster than the thing u mentioned

[ornate elk]

have you updated the webui too?

added git pull?

[hot thorn]

nah, im half dead, just changed webui options

[ornate elk]

then do that please

the dev fixed stuff 2 days ago

[hot thorn]

what will it change?

fixed what?

i updated my venv like today tho

4 times

[ornate elk]

that doesnt help if your not on the updated version

you cant update a venv xD

only manually

[hot thorn]

i mean

delete

make new

delete, repeat

[ornate elk]

yea that doesnt change stuff

you need to be updated on the webui itself

[hot thorn]

so its: set GIT= and then?

[ornate elk]

no

just git pull

in a empty line

[hot thorn]

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --xformers --cuda-stream
where?

oh

in that empty space

[ornate elk]

then save and relaunch

[hot thorn]

if it breaks my SD, when i want to go to bed in few min, im coming to ur house

oh ye, lots of files are being sucked in

i got ur trojan

[ornate elk]

it could break as the reforge dev says ReForge is "Unstable" and users should use Forge Neo xD

[hot thorn]

show mercy

ye, gonna do neo when i wake up
always wanted to know how to dodge bullets just by bending my back

? should i?

oh i see, the pin stuff is for better ppl

as in with better gpus

damn im poor

do i add that malloc in the end?

webui kinda nudges me

[ornate elk]

you can if you want

[hot thorn]

should i?

[ornate elk]

I wouldnt use it cause it could crash in the long run because of it

for a added millisecond of speed its not worth

[hot thorn]

aight, trust

now, to see if u didnt break my SD, and to embrace the bedsheets

oh ye, before i go, is it better use space or underscores in prompts?

[ornate elk]

space

[hot thorn]

why?
i even noticed in sd settings, when u use the auto tag complete stuff, it changes underscore to space as well. doesnt underscore keep 2 words together tho? and prevent bleed?

like when you type wide hips, instead of wide_hips, wont it make more things wide?

[ornate elk]

no it doesnt

underscore just adds a randomness like if you use a . or a !

it doesnt have any function in the webui

the boorutag autocomplete uses underscore to scan for the tags of danbooru but also changes it back to space

[hot thorn]

so... underscore = worse space?

it does have some effect tho no? when chat gpt told me about it, i was going 1 by 1 and checking differences between space and underscore

tho in most cases underscore was worse

ehh either way, sleep time, if i got more questions ill hit u up later

[ornate elk]

Alright cya

[lean totem]

Hi, Someone know macOS? DSDT Patch for NVIDIA egpu on Bootcamp Windows for MacBook Pro 2019 via Thunderbolt

If Someone Can help me, I pay for it, no Problem!

[bold orbit]

sorry im new to all this stuff... i want to gen art but im not sure if unstable diffusion is the best to use or what i should be using... i have a model that im guessing can be used in unstable

[lavish hull]

Automatic 111 and the current version of Python

[ornate elk]

Okay and what's your GPU?

Auto is outdated since a year

[ornate elk]

hey, whats your gpu?

[bold orbit]

Processor 13th Gen Intel(R) Core(TM) i9-13900H, 2600 Mhz, 14 Core(s), 20 Logical Processor(s)

[ornate elk]

GPU (Graphics card) not CPU ^^

[bold orbit]

oh sorry

[bold orbit]

NVIDIA GeForce RTX 4050 Laptop GPU

[ornate elk]

ah okay, then you want to use the setup guides from the pinned messages here.
Follow the Forge Classic guide

With that webui you can then load and use any model that fits in your gpu

[bold orbit]

the The latest Stable Diffusion install Guides for Nvidia and AMD: one?

[ornate elk]

yes

[bold orbit]

ok will do

also sorry whats a webui is that something that comes with it or something i need to download

[ornate elk]

thats something you have to setup, its the programm to use the stable diffusion models

[lone wolf]

Would 9070 XT be a massive downgrade compared to my 7900 XTX? I seem to have issues with mine when it caps on Vram it just is GGWP

Its gonna be difficult to source a 7900 XTX where I live tbh

[burnt salmon]

in comfyui how would i go about controlling the add noise setting in the KSampler. so basically i want my workflow to automatically set the add noise setting to enable if a certain integer equals 0. but disable if its 1 or higher

i have a comparison node set up that outputs a boolean but i cant seem to connect it to the add noise setting

[dusky osprey]

Is there a good build for auto1111 for an AMD gpu? (9060XT) I tried comfy but didn't really liked it with the nodes and stuff

[gusty reef]

nvidia geforce rtx 5070 ti

[ornate elk]

ah okay, and which webui do you use?

[gusty reef]

comfy ui desktop

[ornate elk]

ah okay

then yep its normal

In Forge Neo you can only cap the vram usage

[gusty reef]

Hmmm, I see, in desktop it looks like you set somewhat cap it? In the setting there is a reserved VRam setting which I'm assuming controls how much you want to use/reserve for other tasks.

[ornate elk]

Forge Neo with ROCm or Forge with Zluda, but rocm is currently not stable for rx90 cards

[ornate elk]

yep but that just caps the vram usage not the gpu usage

it will still try to generate as fast as it can

[ornate elk]

its 8gb less vram, so for heavy tasks its a downgrade

like Wan video stuff, but for normal image generation its not that bad

[hot thorn]

@ornate elk u there mate? i woke up xD
in the guide for neo forge, it says to get python 3.11.9, i think I have 3.12,xx something, will it work?

[hot thorn]

@ornate elk pls be alive ;<

[ornate elk]

yes python 3.12 will work too

[hot thorn]

i got some problems after installing.
also how do i get all my extensions i had on reforge?

[ornate elk]

you can reinstall them normaly

[hot thorn]

as in manually? ay,,,

i just linked all my stuff to neo, and tried to generate an image i made today, and got all these in the webui cmd. i assume its not good to see?

[hot thorn]

hmm it might've popped those things because i had open 4 instances of SD from reforge in browser, and they were set to generate forever, but i only had neo open

[hot thorn]

Now i got a different problem. I usually run 4 instances of SD, different neg prompt, and slightly different base promt. When i use neo, it doesnt show me which one is being used right now, doesnt show me previews, and cancelling generation forever.... well doesnt cancel it

[hot thorn]

and it looks like, because canceling generate forever doesnt work, and i have to close the webui manually to stop it, all the open instances of sd just kind of crash
if i open webui again, those already opened instances just dont work, they cant generate, cant load pics in img info, cant refresh wildcards, nothing

neo seems faster, but its wonky af

[bold orbit]

anyone here able to help me for a sec

[paper crag]

got a bit of a conumdrum.
I trained a flux lora to see how it well can turn out, did this on civitai.
After i was done the website didnt allow me to generate with it. Seems to happen if nsfw tags are used in the training,
Downloaded the flux checkpoint on my stable difusion, and its not really working. Say its not allowed.
Any inbuilt safe guard in stable difusion when it comes to realistic lora oor anything like that? (im using zluda)

[bold orbit]

help

[ornate elk]

What are you trying to do?

[vocal burrow]

they're trying to compile numpy

Not sure why tho.

probably chatgpt instructions

[ornate elk]

There is no nsfw filter in the webui itself

[bold orbit]

im trying to run Stable Diffusion

this is what i get when i run the webui-user

[ornate elk]

Which version?
Auto1111, Forge classic, Forge Neo, ?

[bold orbit]

[NVIDIA] Forge Classic

like i was told to

unless i miss understood

[ornate elk]

That should work.
Can you show the complete cmd log when you relaunch the webui-user.bat

[bold orbit]

ill try

do i send it here or dm?

[vocal burrow]

you can send it here there should be no sensible info other than your window user name.

[bold orbit]

oh im not worried about that

just didnt know as its a big text so didnt want to look like spam

oh it did that lol

[vocal burrow]

ok then, don't need to look farther than the first line of log

This program is tested with 3.11.9 Python, but you have 3.14.2.

[ornate elk]

Wrong python version yep

[bold orbit]

ohh do i need it so this all again

this is what i get when i try using the link

[ornate elk]

Nope.
Uninstall python 3.14
Install python 3.11.9 64bit
Delete the venv folder from Forge classic
Relaunch the webui-user.bat

[ornate elk]

Scroll down to windows installer 64bit

[bold orbit]

making sure this is the right one

[ornate elk]

Nope

You need the one 3 up

Windows installer (64bit)

[bold orbit]

oh

see im dumb lol ty

[ornate elk]

Np xD

[bold orbit]

Couldn't launch python

exit code: 9009

stderr:
Python was not found; run without arguments to install from the Microsoft Store, or disable this shortcut from Settings > Apps > Advanced app settings > App execution aliases.

Launch Unsuccessful! Exiting...
Press any key to continue . . .

it said the setup was successful

when i did the python set up

soo idk what im doing wrong

[ornate elk]

Rerun the python installer, select: modify, next, and then checkmark "add python to system environment variable's"

Then relaunch the webui-user.bat

[bold orbit]

ok

[bold orbit]

ok so i think its working but now when i try using a model i get errors

[lone wolf]

I cant really get an XTX... ai pro r9700 tho, is that good for SDXL And can it also perform well in games? I kinda do both, but its very cheap here (Roughly 1000 Euros) so its a bit more than 9070 xt but twice the Vram

[ornate elk]

What's the model file

It has to be at least 2GB to be a model

[bold orbit]

54mb

[ornate elk]

thats not a model

[bold orbit]

im getting them from civitai

ok then idk

thats the info

[ornate elk]

it says Type: Lora, so its a lora

an addon, not a model itself

or try this:
https://civitai.com/models/1986798/cyberrealistic-cyberillustrious-catalyst

[bold orbit]

and a lora is what gives the art the style

[ornate elk]

a lora helps the model to get what the lora is trained on, can be an artstyle, character or something else

[bold orbit]

ah ok

so do i need a model do get the artstyle i want?

[ornate elk]

you need a model to use the lora

it has to be the same Base, in your case Illustrious

[bold orbit]

ok

im so lost im looking for a model but then its giving me a lora

[bold orbit]

ok sooo well this model work with the lora i want to use or no?

i hope im not being to annoying

[hot thorn]

@ornate elk is there anything that can be done with all the junk with neo forge?
with 4 instances open, generate forever only actually generates from 1 instance, and ignores the others
cancel generation forever doesnt do nothing
very often it bugs out and i cant stop a generation
with multiple instances open it doesnt show visual preview.
If i had 1 instance open, closed webui, opened it again, that 1 instance that is already open becomes useless and cannot be interacted with

and thats only stuff i remember on the fly

[ornate elk]

Yes that will work

[ornate elk]

The webuis are not made to be opened multiple times.
But you can always create an issue on forge neos github.
The dev replies fast

[hot thorn]

with reforge i had 16 instances open at one time xD
testing bunch of lora/prompt combinations

btw, does wan on neo forge work well?

[ornate elk]

Only Wan 2.2 16b works on forge neo

Needs a lot of vram but works

[hot thorn]

is that a decent one?
i tried wan sometime ago, with my old gtx 1070, but gave up quickly

is an rtx 3070 12gb enough?

[ornate elk]

Not for 14b

[hot thorn]

oh well, maybe one day then

[hot thorn]

is there anything you would recommend i should use with neo?
i was trying to find some info about what people think about neo, and ive seen something called nunchaku mentioned a lot

[hot thorn]

is FreeU at all usable in SDXL? I installed it, and checked multiple settings, and 0 effect

[bold orbit]

what should i be using?

[hot thorn]

most stable one and used by most people is euler A

[restive sparrow]

Coming from A1111 to ForgeNeo, I've been able to set it up so im getting for the most part 1:1 same results as A1111, but my upscales are sadly lacking in ForgeNeo vs A1111. There a change in contrast/coloring which is not a big deal but the clear difference in crispness/sharpness/fine detail is evident (same seed/model etc. perhaps an option settings?) ForgeNeo has incredible efficiency but that doesnt matter to me vs quality. Is it truly that Neo sacrifices quality for efficency/speed Anyone have any clue? TYVM

[ornate elk]

If you use sage and fast-fp16 then these can sacrifice the quality minimaly

You can then use flash instead

[bold orbit]

hey do you know how you have to use a key word for the model to work... what if i use 2 key works will it then use both or no?

[ornate elk]

only loras sometimes need keywords, models dont.
Check the loras info box on civitai for the keywords

[bold orbit]

sorry thats what i meant was the loras am i able to run them both if i use both keywords?

[deep shuttle]

It looks like the method I used to install Stable Diffusion Automatic1111 no longer works due to recent changes in the repository.
Could someone tell me if there are updated guides available here to install it properly?
My PC specs are:
AMD Ryzen 5 5600G with Radeon Graphics (3.90 GHz)
16 GB of RAM
Any help would be appreciated.

[ornate elk]

do you have only the iGPU of the Ryzen CPU?

no dedicated GPU?

[deep shuttle]

Yes, I only have the integrated GPU (Ryzen 5 5600G). No dedicated GPU.

[ornate elk]

okay thats a bit bad for ai stuff, you probably only can generate using the CPU then.
You can follow the CPU Setup Guide, I updated it right now.

Automatic1111 with CPU Mode:
You need to have Git 64bit and Python 3.10.11 64bit installed.

1. Make a new Folder on your Drive (not on Desktop, Downloads, Documents, Programms, Onedrive) and name it Ai for example:
   *C:\Ai*
2. You go into the folder you created in this case Ai, then click in the File Explorer bar (not searchbar) and type cmd then press enter.
3. Then you copy and paste this command:
   git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git --branch dev
   Press enter and after its done you can close the cmd.
4. Edit the **webui-user.bat **(right click, edit),
   At the line set COMMANDLINE_ARGS= You add: --skip-torch-cuda-test --use-cpu all --no-half --precision full --opt-sub-quad-attention
5. After editing, save and close the bat. Then launch the webui-user.bat
   After the Installation, that can take a while is done, the Webui will autolaunch in your Default Browser.

[deep shuttle]

Okay, I’ll try this.
When I installed it last time, I did it using Anaconda as an intermediary and it worked without any issues, so honestly I don’t know what’s different compared to the standard installation. It even allowed me to run everything from an external drive when I needed to temporarily free up space on my PC.

[ornate elk]

ah okay, yea better use normal python than anaconda.
And yes you can use launch args like:
--ckpt-dir "D:\ExampleModels"
--lora-dir...
to point to the other drives models folder

[deep shuttle]

Thank you very much for your help.
I managed to install it, and with a few modifications I even got it working with the GPU as well. I used Python version 3.10.6, and in the user.bat I added --lowvram --precision full --no-half --skip-torch-cuda-test. So far it’s generating everything correctly, and I’d even say it’s a bit faster—probably because I’m using Python directly instead of Anaconda.

[ornate elk]

perfect, no problem, also try add --opt-sub-quad-attention it can give more performance too

[deep shuttle]

The added flags worked and it’s still generating images for me, but now I’m running into problems with DirectML:

venv "D:\sd\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: v1.10.1-93-gfd68e0c3
Commit hash: fd68e0c3846b07c637c3d57b0c38f06c8485a753
Invalid version: 'onnxruntime-directml'
Warning: Failed to install onnxruntime-directml package, DirectML extension will not work.
Launching Web UI with arguments: --lowvram --precision full --no-half --skip-torch-cuda-test --opt-sub-quad-attention
W0101 14:54:14.642214 14976 venv\Lib\site-packages\torch\distributed\elastic\multiprocessing\redirects.py:29] NOTE: Redirects are currently not supported in Windows or MacOs.
D:\sd\stable-diffusion-webui\venv\lib\site-packages\timm\models\layers_init_.py:48: FutureWarning: Importing from timm.models.layers is deprecated, please import via timm.layers
warnings.warn(f"Importing from {name} is deprecated, please import via timm.layers", FutureWarning)

Although it’s probably related to the fact that I used the same installation link as last time.

[ornate elk]

this is because you set up the nvidia version of the webui and not using directml

you should have followed my setup guide for a proper install for cpu mode

directml wont help you

[deep shuttle]

I followed the exact steps. The only things I changed were the Python version—because for some reason the .bat didn’t recognize that Python was installed on my PC—and the modification of the user.bat to allow me to use the graphics card. Everything else was done exactly as instructed.

I’m installing DirectML so I can use Olive. I was able to use it before, so I honestly don’t know what the problem is now.

But if it is going to be a problem, then I guess I’ll have to settle for just installing ControlNet.
Honestly, as long as I manage to get the generator working properly, I should already be satisfied — I’m just being greedy at this point.
Thanks a lot for being so patient with me.

[ornate elk]

ah okay, so you got directml with olive working in the past?
then you should use the setup guide from the pinned tabs instead of the cpu guide.
You want: Auto1111 with DirectML it supports onnx too
then add --olive to the launch args for olive support if i remember correctly

[restive sparrow]

I was told to and did try without them, but result did not return to quality. I'll try this flash thing. Is just a command line install right

I thought perhaps no half vae absence was the issue as neo doesn't let you do use that.

[restive sparrow]

I've tried disabling sage and Flash and forcing xformers attention but same resulst so it must not be attention stuff

[woeful raven]

guys the automatic1111 zluda is broken, so which ui should i use

?

[lime pivot]

@ornate elk happy new year bro. thanks for being a positive force in this community 🙂

[ornate elk]

Thanks! Happy New year for you too!

[ornate elk]

What is broken? It should still work

[iron lantern]

whats going on here crashes everytime

[woeful raven]

Idk i just read it in your repo thats on fixed messages

in the automatic 1111 zluda part it says that its broken

so i didnt even try iy

btw to use sd i gotta download the pro version of amd software??

what gonna happen with adrenalin

[ornate elk]

oh its fixed, i now removed the Warning sry!

[ornate elk]

no you have to use the normal adrenalin version

[iron lantern]

@ornate elk IS A GOD AND SO HELPFULL !!

[hot thorn]

how do i use FreeU well enough to increase quality without breaking anything?

[versed imp]

i just intall animatediff here, but i cant see the menu,any help please?

[opaque owl]

Does anyone know any easier way to ai image generation, that is easy to use and less coding. I want easy like posting an prompt and bam a photo. Automatic1111 Stable diffusion gave my headache. I want easy like imagine.art, with less restrictions.

Like this here look like alien language.

Don't understand anything, what it means or what it does

[hot thorn]

then SD is not for u man...
SD is the hard part of AI, aka doing everything from scrach
checkpoints, Loras, Embeddings, settings, everything has to be learnt, studied and downloaded

[opaque owl]

Are there service where i can hire people with experience to make ai art for me?

[hot thorn]

mostly patreon i guess?
some people advertise themselves on twitter, and have a link to their patreon.
or people make LORAs for money, not sure about doing everything else xD

and ofc you'd need to find a guy who does style u like to begin with

otherwise i doubt somebody will change everything just to make some pics for some guy on the internet for few bucks

[opaque owl]

Guess will have to wait a few years then to ai gets easier and more powerfull. 2028 then?

[hot thorn]

also depends what kind of art, sfw, nsfw, fantasy, realistic, high detail, cartoony, furry and all that
since all of them need different checkpoints, and depending on checkpoint a different base model, aka different prompts, neg prompts, loras and embeddings

easier with SD? doubt xD
more options maybe

Browser AI and Stable diffusion is like comparing Diablo to path of exile
one is easy and for casuals, the other is complicated af and for min maxing veterans who like to do math to see bigger numbers

[opaque owl]

I use imagine.art browser. Very easy and powerfull, but a lot of restrictions no nsfw allowed there.

[hot thorn]

not a single online generator is gonna allow nsfw, unless there's a high monthly fee to use it

and even then, nsfw ones are low quality

[opaque owl]

So hire people is my only choice then for nsfw?

[hot thorn]

if u dont wanna spend hundreds of hours trying to get what you want? kinda

as a curiosity factoid, ill let you know, i generated about 140 pics in past 2 days, just to check the best negative prompt for what I do xD

same character, same pose, same seed, and just look for minor changes everytime xD

[opaque owl]

I see thank you for the information. ChatGPT wont answer these questions i have. So thank you for the answer.

[hot thorn]

send me a dm, i have some time rn

what questions tho? how to make nsfw content? xD, it wont answer anything nsfw related

[tacit river]

ughughguhg

these ai image generators are so annoying to install, keep meeting problems and problems

[hot thorn]

use a guide from pinned messages, if you follow it exactly its pretty easy, doing it yourself might take some time

[tacit river]

yeah.. im using the guide made by c10

cs1o sorry

[hot thorn]

ye, pretty easy to follow tbh, when i started i was lost several times over too

guide helped a lot, 0 issues now

[tacit river]

im installing forge neo using the guide; edited the webuser bat file and then ran it; I was on the page installing gradio, and BAM was met with a bunch of errors

closed it and retrying it now lol

[tacit river]

yeahh

[hot thorn]

u can always ask chat gpt for help, u can upload screenshots there to ask for specific fixes

[tacit river]

gots it

question ,you said you made a lot of pics just to find a right prompt or setting

do you make money from making ai images?

if you don't mind my asking of course

[hot thorn]

nop, but i plan to

first tho, i need to lock in a nigh perfect combo

then i will go down on it

lets say, my kind of content is popular among people

[tacit river]

ahh

very bad

there r so many people using these generators to make income that way 😭

[tacit river]

I was talking with another person making money that way; but he was making content based on actual people

I went on like a whole ramble about how that's wrong; but it all went in one ear and out the other. Sad to see people use insanely powerful tech in such a way

am I supposed to write the stringe of code like this?

or below the set commandline_args= line?

[hot thorn]

just like on the screenshot u sent

[ornate elk]

Not supported on Forge and forks

[tacit river]

gots it, thanks!

[versed imp]

Forge is inferior to automatic?

[ornate elk]

Send the cmd log here and I can check what's wrong

[tacit river]

Shoot! I closed the command prompt and opened the webuser bat file again

currently it says this

venv "C:\Ai tomfoolery\sd-webui-forge-neo\venv\Scripts\Python.exe"
Python 3.11.9 (tags/v3.11.9:de54cf5, Apr  2 2024, 10:12:12) [MSC v.1938 64 bit (AMD64)]
Version: neo
Installing requirements

[ornate elk]

Looks okay

[tacit river]

Yes, before, it was stuck on a line that said "installing gradio" and afterwards, it just threw a bunch of errrors

[hot thorn]

oh my savior cs1o, tell me man, do u know how to use FreeU?

[tacit river]

cs1o is truly da goat

and also is using da goats pfp

[hot thorn]

he's gandalf

[tacit river]

!!!!!

[ornate elk]

Not much, I just enabled it and played with the sliders a bit.
Its not worth it IMO.
Better use lightning or DMD, or cfg loras

[hot thorn]

ok... wtf are those 3 things u mentioned xDD

[tacit river]

cfg loras... is that the thing where you do the thing and it controls the thingy mabob?

and it n turn makes the thingy ma bob into a thingy ma gog?

[hot thorn]

never heard of cfg loras tbh

[tacit river]

I am an expert in cfg loras

tbh

[hot thorn]

sounds like it

[ornate elk]

Best are lightning loras

2,4,8 step

For sdxl they are a must for details

[tacit river]

hmhm naturally~~~

[hot thorn]

do explain gandalf what even is a lightning lora

and why the fck did i never hear about it, especially when u mention its for details...

[ornate elk]

Its a lora that uses the lightning technique of sd to let you use much less steps for an image.
So instead of 30 steps you can do 8, 4, or even 2.
Or you leave 30 and get insane details xD
https://huggingface.co/ByteDance/SDXL-Lightning/tree/main

[hot thorn]

brah

how dare u keep it hidden from me

and i can just plug it into reforge and neo?

explain the other, explain the other

looks like sleep is off the table for me again...
btw, if i use this lightning stuff, and do 8 steps, i usually do 30 sampling 10 hires, im assuming it will be MUCH faster then?
and if I leave it for the same 30 steps, since I do want that detail, will it be similar speed or much slower?

[ornate elk]

it will be the same speed if you change nothing

it will be faster if you use less steps but the result will be better or same due the lora

[hot thorn]

i want to hug u, and strangle u at the same time right now

i swear man

[ornate elk]

this guy merged the dmd and lightning lora to create a speedup low cfg lora for more details:
https://civitai.com/models/1786270?modelVersionId=2021473

but i like the pure lightning more

[hot thorn]

isnt low cfg kinda bad if i want to make specific characters tho?
since low cfg allows for more freedom in SD

[ornate elk]

depends yea

[hot thorn]

then nah, i wanna keep my 5 cfg

[ornate elk]

v-pred illustrious needs low cfg

so theres that

[ornate elk]

did you get any error now? or all fine?

[hot thorn]

well, i already saved those 3 magic names for future, now im just gonna work on my pos prompt, since im finally done testing EVERYTHING with neg prompt, and hopefully, by tmrw evening i will be a free man once more.

oh yeah, is it true stacking quality tags is redundant?
out of:
best quality, ultra detailed, highres, amazing quality, absurdres, newest, which would u keep?

[tacit river]

it's still saying installing requirements

taking a while lol

maybe its stuck?

[hot thorn]

those are long

probably its creating venv

[tacit river]

ahhh

yeah even on task manager, my cpu is not above 10% usage

I'm guessing that's a bad thing?

[hot thorn]

?
if its downloading and installing a bunch of small files then it wont go to crazy numbers man

[tacit river]

oh okay

I asked chatgpt and it said to check that, and if cpu usage is below 10%, it mean bad

[hot thorn]

chat gpt is often spouting bullshit tbh

i think it usually takes me about 10 min maybe 15 to install everything

if i wasnt up for 16h now with still much to do, id help out, but im dying, so i only sometimes check the chat, and answer easy stuff

trying to take a bit of weight off of cs1o mighty shoulders, to pay him back for all help he provided

[tacit river]

ahhh

thank u for da help regardless!

much much appreciated

it's been like an hour lol

and everything else has been instaleld

just that it crashed at the installing gradio line,

[ornate elk]

No, it can take a while

[hot thorn]

an hour for the requirements?

thats kinda sus

[ornate elk]

@tacit river How much RAM do you have and what's your CPU?

[tacit river]

i have 32 gigs and 9900x

ryzen 9 9900x

good sir

[ornate elk]

Then it should install much faster

[tacit river]

should i close it and reopen it then?

[ornate elk]

What does it show currently?

[tacit river]

Python 3.11.9 (tags/v3.11.9:de54cf5, Apr  2 2024, 10:12:12) [MSC v.1938 64 bit (AMD64)]
Version: neo
Installing requirements

it says this

like only this

[ornate elk]

Can you screenshot the cmd?

The whole window of cmd

[tacit river]

this is just it lol

[ornate elk]

Very strange

What's your GPU?

[tacit river]

RTX 5070

i dont think its a system issue

something else probably

[ornate elk]

Okay then close the cmd.
Delete the venv folder and relaunch the webui-user.bat

[tacit river]

okay boss

okay nows its installing everything again

and taking time again on the installing requirements part

[ornate elk]

Okay make sure you also have enough free space on C

[tacit river]

I have 65 gigs free

64**

[tacit river]

uhhh, it's still stuck on the line installing requirements

lol

OMG!

IT WOKEEE!

THANK U CS1O FOR THE GREAT GUIDE AND HELP! AND NIHIRE

ugh

now it says i need a model

i put a model into the stable diffusion folder (model -> stable diffusion

but im just getting an empty picture; nothing else

uhhh wat is dat?

for wat lol?

im sure this is some sort of issue that can be fixed with some tweaking haha

[ornate elk]

Scam dont click!

[ornate elk]

You have to click on XL at the top left

Then try to generate an image

[tacit river]

yes sir!

I clicked on it

and then loooked on the console log and this is what appeared

  x_sample = x_sample.astype(np.uint8)

[tacit river]

phew! I copied the link and pasted it into virustotal; it was a virus link unfortunately

[ornate elk]

Can you try this model instead?
https://civitai.com/models/376130?modelVersionId=2553832

With default settings on XL preset

[tacit river]

of course!

I'll give it a shopt

Thank you very much!

It'll take a while to download so,

[ornate elk]

Ok

Can you show the other model file

Which you tried before

[tacit river]

Yes!

this is the file

I downloaded it from here
https://huggingface.co/spaces/Asahina2K/animagine-xl-4.0

[ornate elk]

Hmm that should work too.

[tacit river]

yes

but it isn't working, just generates a black image

[ornate elk]

Have you tested the other one?

[tacit river]

It's still downloading

[ornate elk]

Ah can you relaunch the webui and then select XL at the top left then press generate and prompt for just Cute Cat

It should work

[tacit river]

://///

[ornate elk]

Oh you have to type something into the negative prompt too
For example blurry

Then it should work

The dev of forge neo mentioned that a few days ago

[tacit river]

waaaattt

YOOOOOO

IT WORKS!!!!!!!

THANK YOU SO MUCH CS1O!!!!

UR THE BESTTTT!!!!

[ornate elk]

Perfect no problem 🙂

[tacit river]

uhmmm

is it supposed to take long?

it's takng like a minute to generate a prompt

[ornate elk]

It should take seconds

Can you share the complete cmd log?

[tacit river]

[Unload] Trying to free 5079.41 MB for cuda:0 with 0 models keep loaded ... Done.██████| 20/20 [00:04<00:00,  4.67it/s]
[Memory Management] Target: IntegratedAutoencoderKL, Free GPU: 5908.97 MB, Model Require: 159.56 MB, Previously Loaded: 0.00 MB, Inference Require: 2238.40 MB, Remaining: 3511.01 MB, Moving model(s) has taken 0.03 seconds
Total progress: 100%|██████████████████████████████████████████████████████████████████| 20/20 [00:05<00:00,  4.00it/s]

huh, now it's fast...

[ornate elk]

Seconds fast?

[tacit river]

yup

[ornate elk]

Perfect then all is good

[tacit river]

Thank you so much bro!

helped me out a tonnnn

been struggling with this for a while now

[ornate elk]

No problem ^^

[tacit river]

^^

[grizzled mural]

is there a way i can get reForge Inpaint tab style on normal Forge? (I prefer reForge's inpainting workflow)

[hot thorn]

do my generations change after deleting venv and creating a new one?
i tried recreating my pic from 4 days ago, and its quite a bit different.

[onyx solar]

Recreating a venv itself wouldn't necessarily do that by default, but if you had anything involved that is different when you try to recreate the image at all, including any LoRAs, seed value, CFG, or anything at all, really, then you will likely get different results.

[hot thorn]

i used png info man, its exactly the same, loras, embeddings, checkpoints everything

[hot thorn]

i swear, chat gpt is such a damn garbage when it comes to stable diffusion.
asking it to fix 1 problem, follwing instructions, boom 20 different problems appear, and you gotta spend an hour fixing them back

[vocal burrow]

It shouldn't.

[hot thorn]

already fixed it, cs1o told me to update my stuff, and the updates made my stuff look worse

so i rolled back

[hot thorn]

@ornate elk u there?

https://civitai.com/models/1786270?modelVersionId=2021473
https://civitai.com/models/350450?modelVersionId=391994
i found these 2, are they decent for lightning?

and how do i use it properly

[ornate elk]

i shared these two yesterday xD they are okay but the first one requires low cfg

[hot thorn]

ok, then if i use the 2nd one, which step version should I use?
and then how many steps do i set up, can i keep my 5cfg?
do I still use hires? adetailers?

[hot thorn]

;<

[ornate elk]

Yes keep everything the same

You can test between the 3 versions

[hot thorn]

everything the same? arent i supposed to lower steps?
also chat gpt said to not use adetailers and hires pass at all
wont it take longer if i keep the same settings?
sampling steps 30, hires 10, hires fix 1.5, starting resolution 1080, cfg 5, denoising 0.36
@ornate elk

[ornate elk]

you have to test around a bit

also test different lora strenghts

[woeful raven]

yo i just installed auto1111 and it says "Compilation in progress. Please wait..." in cmd

what should i do

[ornate elk]

now you only need to wait, it will take 15-30mins

then its compiled, first image could take also a bit longer

[woeful raven]

what will show on cmd when compiled?

And btw that compilation gonna happen for every model that i use?

[ornate elk]

nope only first time if the model is different like sd 1.5 or sdxl, flux etc

[ornate elk]

nothing, try to generate one image and when its done its finished with compiling

[woeful raven]

so i just let it like that and wait?

its compiling for about 30 mins now

[ornate elk]

yep just wait it will work

[woeful raven]

k, btw before this launch that im waiting i closed it when it was compiling, do u think this could break anything?

[ornate elk]

no it will compile again

[woeful raven]

k now its working, how do i check if its using gpu

[ornate elk]

check the vram usage in task manager

and if the image takes seconds instead of minutes its using gpu

[woeful raven]

im downloading a checkpoint but its for sdxl, wth is that

if i download and put it in models folder it gonna work

?

[ornate elk]

yes it will work

[light maple]

Hey I'm trying to use a new stable diffusion webui, I was using a legacy version of it before but after getting a new graphics card or updating the tool I don't seem to be able to use it anymore

So I installed a newer SD webui but it seems to have hidden or modified some of the settings in the UI. I tried generating an image using the same settings that I could find, the generation works but it still comes out strange or messy and gives me a much stranger result even when using the exact same prompt I had before . I had to search to find the clip step setting. But I'm wondering if there are any other default values under the hood that could be different. I know this is a bit of a board question here but could anyone help me with that?

[ornate elk]

hey, whats your gpu and which webui have you setup?

[fossil meteor]

Hey, all. I'm very new to this platform and I'm trying to install the webui. Can anyone point me to a step-by-step guide that doesn't assume I'm familiar with the process?

I think I've got most of it settled, but i have python 3.13 installed and it's jacking up my process. Not sure where to proceed. Thanks!

[ornate elk]

hey you need to uninstall python 3.13

its not supprted

and whats your gpu?

[light maple]

Using a 5080 now (was using a 3060 Ti before)

And I'm attempting to use Stable Diffusion WebUI Forge - Classic

When I tried to use the Legacy version I had installed I get this error now:

AcceleratorError: CUDA error: no kernel image is available for execution on the device Search for `cudaErrorNoKernelImageForDevice' in https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__TYPES.html for more information. CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect. For debugging consider passing CUDA_LAUNCH_BLOCKING=1 Compile with `TORCH_USE_CUDA_DSA` to enable device-side assertions.

[ornate elk]

can you try to install Forge Neo instead? its the best right now for rtx50 gpus

the setup guide is in the pinned messages of this channel

[waxen ivy]

@ornate elk Have you tried the StabilityMatrix setup from their web page? I installed it with the mobile options, download several UIs to try, and settled for the one I prefer. When you start it it also tells you if there are updates and will install any updates for you. Gives links for downloading models as well.

[ornate elk]

yes i know stabiliy matrix but i dont recommend it as it adds a layer where thing can fail or cause problems

[marble moth]

Hi, I am currently trying to install "Insightface" but keep getting the same errors for days now. Can someone please help me with it?

Here are the error code

For context, it is via ComfyUI. Thank you in advance for the support

[iron lantern]

non stop crashig any clue ?

[iron lantern]

had to uninstall forgeui and now iam stuck getting it back installed

[gloomy bolt]

@ornate elk I want to try ZIT that everyone's been looking at https://civitai.com/models/2169712/z-image-turbo-quantized-for-low-vram?modelVersionId=2549032 but it looks completely different from what i've been using with forge and illust. models so i'm not sure ik how to overhaul my stuff

[ornate elk]

Are you on Forge Neo?

[gloomy bolt]

how can i check if i'm on regular forge or neo?

[ornate elk]

When the webui is open scroll to the bottom

There it shows the version

[gloomy bolt]

https://github.com/lllyasviel/stable-diffusion-webui-forge/commit/f53307881bfd824dbdce6ac0d4bba04d9a74ab36

looks like regular forgeui

[ornate elk]

Okay yep

Then z image won't work.
As only Forge Neo and Comfyui support it

[gloomy bolt]

Am i able to make the switch? Forge Neo is the same as ForgeUI other than it supporting more models right

[ornate elk]

Yea its more updated and less bloated then Forge

And its still maintained and has the best performance boosts integrated

[gloomy bolt]

hm awesome. R u able to DM me and guide me through the process? Not really good with the technical stuff. I'm heading off rn but will be back later

[ornate elk]

you can find the Guide here:
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides

[gloomy bolt]

Will this create a new branch that links my current forge & auto1111 files? I have a bunch of loras and stuff in my current folder

[ornate elk]

its a new seperate folder and you can then link your models over to forge neo, its also in the guide how to do that

[woeful raven]

is there anyway to optimize sd for amd and make it faster?

its slow if using some models

[ornate elk]

whats your gpu and webui? and which models are slow

[woeful raven]

rx 7600, auto1111, pony diffusion is a little slow, took like 6 mins to generate an 512x512 image with 30 steps

[ornate elk]

thats not normal, can you show your used settings in txt2img and your cmd log?

[woeful raven]

cmd log after generating or just when i launch?

[ornate elk]

after generating

[cosmic furnace]

cs1o ) howdy )

any updates on the rock ?

[ornate elk]

sage attention works on comfyui, and maybe flash attention too

[cosmic furnace]

also . if yur familiar with z-image why this image of txt2img 1256-1256

feels not as good by quality as this img2img direct upscale to 1400-1400

original txt2img feels kinda... pixelated =\

[ornate elk]

yea z-image quality is always a bit pixelated for me too

its because its a turbo with cfg 1

[cosmic furnace]

then its better speed but worse quality then flux? me & friend kinda starting to feel like it

yet again just direct txt2image her face feels pixelated a lil bit

[woeful raven]

well it gonna take a while

[cosmic furnace]

pretty sure img2img uspcale without any highres fix gave better results 🤔

[ornate elk]

okay interrupt that and send me a screenshot of the settings and the cmd log

[ornate elk]

but you used a larger res, her face is very small in the initial image

less pixels = less quality

[woeful raven]

like that?

[cosmic furnace]

i tried upstraight txt2img 1400-1400 and first txt2img 1256-1256 . and then this image into 1400-1400 in img2img tab, and it feels when in img2img conversion ended into better quality. am i imagining things or maybe thats how u can get better quality in z-image? 🙂

can u test it later?)

[ornate elk]

you added launch args that slow everything down, thats why its slow.
Remove --no-half --precision full --opt-sdp-attention --medvram from the webui-user.bat

[ornate elk]

i dont get it sry xD maybe send the txt2img and img2img settings to compare too

[cosmic furnace]

txt2img

img2img

[woeful raven]

did it, its faster now but still slow

ETA is like 2 mins

[cosmic furnace]

let me say this again. lol , it feels that straight up txt2img generation on 1400-1400 res, gives more bad-pixelated image

[ornate elk]

then add --medvram-sdxl to the launch args and try again

[cosmic furnace]

than when u do 1256-1256 in txt2img first & then just send it into img2img & generate 1400-1400 img from it.

can that be true ? or im imagining things?

also it feels that my friends z-image he uses on 1060 . kinda giving better detailed face =\ .im using this one zimage

[woeful raven]

now its 1 min

[ornate elk]

it can be that forge with zluda is faster for your gpu instead of auto1111 with zluda.

[woeful raven]

so should i install forgre

forge

?

[ornate elk]

you can try, and then compare, because on gpus with less vram its mostly faster

[woeful raven]

k

[ornate elk]

i have that one too

[cosmic furnace]

auto111 is kinda old already =\ . ofc forge shud be better +)

[cosmic furnace]

and its okay ? have u tried other ones?)

[ornate elk]

yea thats okay

[cosmic furnace]

i mean... its kinda paradox that friend with 1060 have better quality. it is kinda feels like it. I doubt i have more "optimized" cut quality version than him. he used face adetailer but when i try to inpaint face it doesnt feels it do anything =\

[ornate elk]

for me adetailer works on zimage

[cosmic furnace]

yeah but i prefer to just use inpaint on faces =\

its kinda same . isnt it ?)

[cosmic furnace]

oh wait. i wasnt using only masked. just whole image xD

now i guess really face alot better)

mb zimage works better when in img2img than just from text tho ? 🤨

[cosmic furnace]

U wanna say theres something can be done about it ? have u tried different zimages? mb theres one lil bit slower but alot better quality ?)

[ornate elk]

no i dont have the time to test multiple but i tested the fp8 and the allinone and boath gave nearly the same result

[cosmic furnace]

is there something u recommend trying tho ?)

to increase quality or u think the speed worth it?

[livid bridge]

Hi everyone! I'm trying to get SD installed and I've been running into an issue:
"remote: Repository not found.
fatal: repository 'https://github.com/Stability-AI/stablediffusion.git/' not found"
I've googled and searched the web but I'm too smoothbrained to know which repo is currently the most up to date. I think I read the official repo hasn't been updated since 2022.

[woeful raven]

follow the steps in the pinned messages

in the first pinned message

i mean

[fossil meteor]

The laptop I'm using doesn't have a dedicated GPU; it's a little dated. Ryzen 5 4500U. I do have another laptop with a dedicated 1650 in it, but that's also dated.

[ornate elk]

Hmm then you can try setup AMD Automatic1111 with DirectML

The 1650 can use Nvidia Forge setup guide

[vast crest]

Hi, is there anyone using Google Colab for LORA training? I need some advice.

[ornate elk]

hey, you have to use the updated guides from the pinned messages to install SD

whats your gpu btw? you might want to look at Forge Neo as its still maintained

[ornate elk]

@cosmic furnace tested the txt2img 1256 then to img2img with 1400.
Without adetailer.
Idk but boath look good

[gloomy bolt]

i'm following the guide and also how i linked auto1111 to my past forge

does the reference look good or is it supposed to go on another line

[ornate elk]

you have to change the / to \ in your path

the rest looks good

[gloomy bolt]

is this for lora's or will it link the whole thing?

[ornate elk]

it will link the whole models folder and embeddings, loras etc

[gloomy bolt]

awesome

i just save and run the webui right

[gloomy bolt]

My loras, models, etc. doesn't seem to be linked

this is what I had to run on my old ForgeUI, but idk how much of it applies here

[ornate elk]

but did the models worked?

only the loras missing?

[gloomy bolt]

no both are missing

TIs too

this is the one for forge neo

[ornate elk]

can you show me the cmd log when you launch the webui-user.bat?

[ornate elk]

ah okay the path in there looks broken,
you have to use the " " around the path in the webui-user.bat

[gloomy bolt]

ah like this?

[ornate elk]

save and relaunch

[gloomy bolt]

👍

am i able to delete my old forge or will that mess with things

i think the new forge is just linked with auto1111 not the old forge

[ornate elk]

yep you can delete the old forge, but save your output folder

[gloomy bolt]

Ok, the loras and TI loaded in

still missing the models though

[ornate elk]

can you show the cmd log again?

[ornate elk]

hmm it looks correct

can you show how it looks in the webui?

[ornate elk]

can you click into the checkpoint drop down

at the top

[gloomy bolt]

o ok those are there

[ornate elk]

ok good

select a preset and a model that fits and then the loras will show up if they are compatible

[gloomy bolt]

ah ok

so everything on neoforge works the same as the forgeui right

[ornate elk]

not everything but mostly yes, the dev of forge neo removed some bloat stuff and some uneeded functions

[gloomy bolt]

ok nice

[ornate elk]

this stuff got removed

[gloomy bolt]

mm ok

[ornate elk]

i found nothing missing for me in forge neo

[gloomy bolt]

yea i don't think i use most of those

[ornate elk]

adetailer extension works too

[gloomy bolt]

i'm trying to put the launcher onto my desktop like I did with the forgeui but don't remember lol

[ornate elk]

right click the webui-user.bat, click more options at the bottom, click "Send to", select Desktop

[gloomy bolt]

nice ty

[ornate elk]

np

[gloomy bolt]

I have a question about one of the extensions I was using, but let me add all the extensions on first

I can remove install from webui-user right

[ornate elk]

install ?

what are you trying to do

[gloomy bolt]

there was the "install" line here

[ornate elk]

yea remove that

its not there by default

[gloomy bolt]

interesting

Ok, so I had a question about the ExtraSchedulers extension a friend helped me install last time. He said it works great for anime related images, but if I want to disable it for like realisim, do I just delete the line "m + (M-m)*(1-x)**3"?

[ornate elk]

no if you want to disable them just dont select them as the scheduler

[gloomy bolt]

mm so i just disable this?

[ornate elk]

no i mean just dont select the extra scheduler then

in the dropdown

for realism just set scheduler to automatic or choose an other one

[gloomy bolt]

wait nvm

tripping i see the new schedulers

so it turns out I was using Euler CFG++ which is a new sampler that was added by the extension, but the scheduler I was using was "simple" which I believe is there by default. Does that mean I had it disabled the whole time or something

also this extension is not showing up on my tabs even after applying and restarting UI

[ornate elk]

No that means you just used a new sampler

[gloomy bolt]

mm ok

[ornate elk]

That civitai extension is not updated since 1-2 years.
On github even 3. It wont work on newer webuis like neo.
But there is a similar extension which is still maintained:
https://github.com/anxety-solo/sd-civitai-browser-plus

You have to delete the other one before Installing

[gloomy bolt]

Awesome works now ty

one more set up question sry 😭

Is the upscaler "r-esrgan 4x+ anime6b" something that was added manually or was it removed from Neo?

[ornate elk]

It was not added in forge neo.
You can download it from here:
https://openmodeldb.info/models/4x-realesrgan-x4plus-anime-6b

It goes in the Esrgan folder then restart the webui to see it.

[gloomy bolt]

Ok restarted webui and it's not showing up

actually let me put it in the auto111 file

[ornate elk]

You have to add it to your linked Auto model folder

There in esrgan

[gloomy bolt]

alr works 😭

mm i think that's everything for setup, anything else you recc to change or add

like the webui, I don't need xformers and stuff anymore right

[ornate elk]

Nope your fine.
For Illustrious anime images I always recommed the Booru Tag Autocomplete extension. Its really good for prompting in booru tag style.
And forge couple extension if you want two different characters in the same image

[gloomy bolt]

mm ok let me take a look into those

o ok the autocomplete extension is really nice

ok so for ZIT now 😭 ... for

https://civitai.com/models/2169712?modelVersionId=2549032
I need to install

1. https://civitai.com/models/2169712?modelVersionId=2549032 (the model itself)
2. https://civitai.com/models/2169712?modelVersionId=2474529 (text_encoders (?))
3. https://civitai.com/models/2168935?modelVersionId=2442479 (VAE)

there's no text encoders (?)

[ornate elk]

Create one and call it text_encoders or text_encoder
Idk if it was with s or not at the end

[gloomy bolt]

i think it's w/ s

downloading still

[ornate elk]

I'm off now gn

[gloomy bolt]

gn ty for helping

[ornate elk]

No problem!

[iron lantern]

RuntimeError: CUDA error: an illegal instruction was encountered
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1.
Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.

why am i getting this

[iron lantern]

ok so when i run large batch size it crashes ... i used to be able to run 100 on a 3060 12gb no problem once installng comfyui my pc has gone to poop lol

[hot thorn]

@ornate elk mate, could you take a loot at something for me?

worst quality, low quality, worst detail, lowres, sketch, censor, simple background, error, jpeg artifacts, signature, watermark, username, moiré pattern, down sampling, distorted, blurry, poorly drawn, low-resolution, pixelated, fake, semi-realistic, duplicate, SmoothNegative_Hands-neg, bad anatomy, extra limbs, extra digits, fewer digits, deformed hands, heart-shaped pupils, cgi, 3d, glitch
would you delete anything from this neg prompt?

I recently did a lot of changes, and ended up with all of these till the embedding, but noticed a bunch of problems, so i added all that u can see past the embedding back to the prompt

I still cut off a few

this was the original: worst quality, low quality, worst detail, lowres, sketch, censor, simple background, error, jpeg artifacts, signature, watermark, username, moiré pattern, down sampling, distorted, blurry, poorly drawn, low-resolution, duplicate, pixelated, fake, glitch, semi-realistic, cgi, 3d, (cape), heart-shaped pupils, (futa, futanari), see-through hair, transparent hair, bad anatomy, disconnected limbs, deformed hands, extra limbs, HDA_BadHands_neg, FastNegative, extra digits, fewer digits,

[woeful raven]

@ornate elk i downloaded forge but its slower than auto111

theres osmething wrong

[ornate elk]

hmm okay strange

[ornate elk]

looks okay to me, is it for realisitc images?, then you can also add Anime or cartoon in there too

[hot thorn]

nah nah, its for anime specifically

[ornate elk]

ah okay

[hot thorn]

for anime its fine? should i put back any of the stuff from original prompt?

[ornate elk]

no i think thats enough

[jagged crane]

How can I start training the Stable Diffusion model?

[cosmic furnace]

tried to update forge & now it doesnt run o.0

[ornate elk]

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

[ornate elk]

Can you show the complete cmd log?

[ornate elk]

Two issues:
ROCm agent enumerator failed: [WinError 267] Неверно задано имя папки
ROCm: version=Python311
ZLUDA support: experimental
ZLUDA load: path='G:\SD\SD - Forge-Webui-ZLUD\333\stable-diffusion-webui-amdgpu-forge.zluda' nightly=False
Failed to load ZLUDA: Could not find module 'G:\SD\Python311\bin\rocblas.dll' (or one of its dependencies). Try using the full path with constructor syntax.
Using CPU-only torch

Did something in your paths changed?
It says rocm = python311 that's very broken

Check your system environment path variables

[cosmic furnace]

seems fine o.0 . tried to update forge with git pull -_-

not sure if its the case =\

[ornate elk]

Go into that path setting

Check for a path that points to rocm/bin

[ornate elk]

Looks okay

Need to go off now its very late.

[cosmic furnace]

ill try reboot & reinstall venv-zluda folders i guess =\

weird. doubt its cause i launched forge neo before

[gloomy bolt]

hello, the CivitAI Browser+ https://github.com/anxety-solo/sd-civitai-browser-plus.git is not updating the model previews, I think it's scanning the wrong folder or something how can I resolve this?

[solar hamlet]

hey guys thanks for the invite. unable to install SD it told me to disable app execution aliases on python and python3, i did so, try to run webui-user and now saying it cant find python at all

INFO: Could not find files for the given pattern(s). Couldn't launch python exit code: 9009 stderr: '-c' is not recognized as an internal or external command, operable program or batch file.

all right, going to uninstall and reinstall and try again

[solar hamlet]

I still can’t figure this out

It says I should see a thing called webui when I go to this one url. I pasted it in, and it claims it can’t connect

[lucid yoke]

Python3 is not really needed, " install sd" what? i assume your trying to install a outdated webui

try checking the pinned messages for the guide of CS1o

and do not use chatgpt

[solar hamlet]

Will do, thank you! I can show you the tutorial I first tried and got the farthest with

[lucid yoke]

honestly installing a webui should not be more then launching a .bat file or at worst installing python 3.10

just python 3/python is vague

most youtube guides are frankly bad or outdated

im on SwarmUI

[solar hamlet]

https://stable-diffusion-art.com/install-windows/

I tried this one

This is where i started. when it didnt work, I tried to troubleshoot it and from there i just started trying one thing after another.

it's frustrating when you follow a tutorial perfectly as they say and you still get errors, and the tutorial doesnt even mention them as a possibility

And it wasn't just one error it was many because i tried the tweaks they suggested

[lucid yoke]

oh damn seems outdated since its uses a1111

[solar hamlet]

said this right on the page

[lucid yoke]

a1111 is outdated

[solar hamlet]

I checked before starting to see that it was a new tutorial

Thanks for letting me know. I had no way to know, I'm new to this

[lucid yoke]

a1111 still works yes but good luck trying to do the newer stuff on it

[solar hamlet]

well that explains why nothing was working when i tried stability matrix

[lucid yoke]

if you wanna mess with like z-image and the like i recommend swarmUI since im on that but it depends on your hardware

what gpu do you have?

[solar hamlet]

lenovo thinkpad, nvidia geforce rfx 4070 (i think)

[lucid yoke]

hmm if its a 4070 that should be fine yeah

https://github.com/mcmonkeyprojects/SwarmUI?tab=readme-ov-file#installing-on-windows
its a one click install if you match the requirements

if your on win10 that is

if your on win11 it should be fine

@solar hamlet hope it works for you but if you run into a issue do lmk or hop into the swarmUI discord, its a bit more active and the dev is there too

[heady pine]

Could someone show me a screenshot of what it looks like when you are applying a style? I would like to apply one to some images, but I have no real idea how to do it.

[lucid yoke]

You mean image to image with a lora/style

Initial image + 0.5 creativity or mess with it a little till your happy with the results

[solar hamlet]

Thank you so much and I will give it a shot tomorrow when I’m back awake.

[jagged crane]

help

[lucid yoke]

im going to be honest and say i have no idea what you mean, like the main checkpoint, a lora?

you wanna add a style or a new character?

certain concepts?

[heady pine]

txt to image, I want to use a hyperrealistic style and see what results I get, but I am unsure how to apply the style. - I am using an anime checkpoint.

[lucid yoke]

well depending on the UI

bascially this

[heady pine]

I appreciate your video, but just a final question. I am trying to apply this style:

https://civitai.com/models/626603/hyperrealistic-pony-or-illustrious?modelVersionId=1914557

For the positive prompt, I won't need to add this: lora:Hyperrealistic_illustrious:1

I will only need to add the trigger word, which is this: Hype4realistic

Is that correct?

Sorry, but I have been wondering this for a while

[lucid yoke]

it definitely depends on the ui you are using

[heady pine]

Ok, thanks for your help on this

[lucid yoke]

on forge/a1111? yes you need to add that <lora> part

on swarmUI? as my video

on comfyUI? add the node and trigger word

[heady pine]

I use A1111, so with what you have said, now I know I will need to add the <lora> part. Thank you.

[late cipher]

I just installed stable diffusion how do fix this?

Can anyone help?

[vocal burrow]

send logs

[late cipher]

Wdym

@vocal burrow

[vocal burrow]

can t figure out what s wrong with just that screenshot

need to see what the logs say.

[late cipher]

uhh how do i send the logs?

[vocal burrow]

the white text in the console

[late cipher]

I was trying to generate an image and it js said error connection error

[vocal burrow]

copy paste it in .txt, drop said .txt here

[late cipher]

ur talking about this?

[vocal burrow]

yes, send the whole thing

[late cipher]

Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA GeForce GTX 1650 : native
VAE dtype preferences: [torch.float32] -> torch.float32
CUDA Using Stream: False
D:\stablediffusion\system\python\lib\site-packages\transformers\utils\hub.py:128: FutureWarning: Using TRANSFORMERS_CACHE is deprecated and will be removed in v5 of Transformers. Use HF_HOME instead.
warnings.warn(
Using xformers cross attention
Using xformers attention for VAE
ControlNet preprocessor location: D:\stablediffusion\webui\models\ControlNetPreprocessor
2026-01-06 20:27:02,544 - ControlNet - INFO - ControlNet UI callback registered.
Model selected: {'checkpoint_info': {'filename': 'D:\stablediffusion\webui\models\Stable-diffusion\ponyDiffusionV6XL_v6StartWithThisOne.safetensors', 'hash': 'e577480d'}, 'additional_modules': [], 'unet_storage_dtype': None}
Using online LoRAs in FP16: False
Running on local URL: http://127.0.0.1:7860

To create a public link, set share=True in launch().
Startup time: 20.9s (prepare environment: 4.1s, launcher: 0.5s, import torch: 9.1s, initialize shared: 0.2s, other imports: 0.4s, load scripts: 2.8s, create ui: 2.4s, gradio launch: 1.3s).
Environment vars changed: {'stream': False, 'inference_memory': 1024.0, 'pin_shared_memory': False}
[GPU Setting] You will use 74.99% GPU memory (3071.00 MB) to load weights, and use 25.01% GPU memory (1024.00 MB) to do matrix computation.
Loading Model: {'checkpoint_info': {'filename': 'D:\stablediffusion\webui\models\Stable-diffusion\ponyDiffusionV6XL_v6StartWithThisOne.safetensors', 'hash': 'e577480d'}, 'additional_modules': [], 'unet_storage_dtype': None}
[Unload] Trying to free all memory for cuda:0 with 0 models keep loaded ... Done.
Press any key to continue . . .

This is what itsays

and here I tried to make another one but same error

[late cipher]

Uhh so what do i do?

[vocal burrow]

make sure you ve white listed http://127.0.0.1:7860/ in your adblocker if you use one

also which forge are you using ?

[late cipher]

i js installed the latest one

[late cipher]

oh like turn of my ad blocker for that?

[vocal burrow]

just add the url to the whitelist

or turn it off sure if you don t know how to whitelist

[late cipher]

Well im using Ublockl

Ublock origins

[vocal burrow]

and regarding forge ? which one are you using ?

[late cipher]

Im using this one

[vocal burrow]

forge ? forge classic ? forge neo ?

[vocal burrow]

that does nt tell me

[late cipher]

well i used a one click installation path so im not sure which one im using but im js using forge ig

since i installed it from here

[vocal burrow]

what url

[late cipher]

https://github.com/lllyasviel/stable-diffusion-webui-forge?tab=readme-ov-file

this one

[vocal burrow]

ok there we go, so the standard forge

[late cipher]

yes

[vocal burrow]

you dont close the console after opening the webpage, do you.

[late cipher]

uhm

is it bad if i do

[vocal burrow]

yes... you re closing the whole server by doing so.

[late cipher]

i click the run.bat thing

and then it runs and opens the webpage

[vocal burrow]

and then you DO NOT close it untill you re done generating stuff

[late cipher]

Okay i js started it now

Welp uh

i tried generating even with the console

on

same problem

[vocal burrow]

ok

[late cipher]

plus whitelisted the webpage with my ad blocker

soo what do i do?

[vocal burrow]

updated gpu driver ? are you on a laptop ?

[late cipher]

on a computer

pc

[vocal burrow]

try using forge classic instead https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides#nvidia-forge-neo

[late cipher]

do i uninstall the one i ahve

have*]

[vocal burrow]

just install it somewhere else

[late cipher]

uhh so how do i install this one?

[vocal burrow]

I just sent you a link to guide explaining how to install it

[late cipher]

it says this

@vocal burrow

[vocal burrow]

line 5 of the guide

[late cipher]

yeah i have it installed

[vocal burrow]

did you check the box ?

[late cipher]

wdym

what box

[vocal burrow]

like it says, did you check the "Add Python to Path" during the installation ?

[late cipher]

ohh

no i didntrr

ill reinstall

yeah i checked it this time still the same problem

[ornate elk]

Thats python 3.10 not 3.11

[late cipher]

huih

then what do i do

[ornate elk]

Uninstall python 3.10 and Install 3.11.9

[late cipher]

done

well

still teh same thing

[vocal burrow]

and you checked the box when installing 3.11 ?

[late cipher]

yes

i can uninstall it and install it again

[vocal burrow]

What do where python, python --version and py --list-paths say ? (run those in cmd)

[late cipher]

this

[vocal burrow]

ok looks good

what do logs say when you run webui-user.bat ?

[late cipher]

same as this

[vocal burrow]

if there is a venv folder next to the .bat delete it (the venv folder) then relaunch it afterwards

[late cipher]

well

there isnt any venv thats the thing

[vocal burrow]

ok please take a screenshot of it again, just to make sure

[vocal burrow]

where did you install forge classic ?

[late cipher]

local disk C

[vocal burrow]

ok but where on c ?

[vocal burrow]

ok

show the content of webui-user.bat please (right click -> edit it)

[vocal burrow]

ok... it might be windows being windows again.

In the start menu, search for systempropertiesadvanced and run it then select "environment variables" (should be at the bottom).
Then doubleclick the PATH variable for the system (not for the user) and screenshot its content.

[vocal burrow]

ok, bump the python entry (the last one) to the top

reboot

then launch webui-user.bat again

[late cipher]

wdym by bumping

[vocal burrow]

place it to the top of the list using the "Move Up" button

[late cipher]

done

[vocal burrow]

reboot, relaunch webui-user.bat

[late cipher]

do i restart my computer

[vocal burrow]

yes

[late cipher]

im back

still says teh same thing

[vocal burrow]

so weird

what s in this folder ?

[vocal burrow]

ok edit the webui-user.bat file again.
replace the
:: set PYTHON= line by
set PYTHON=xxxxx xxxxx being the path of the python.exe shown in your latest screenshot

[late cipher]

im assuming this?

[vocal burrow]

replace xxxxx with the path of the python.exe

and remove the :: at the beginning of the line.

[late cipher]

like this?

[vocal burrow]

the full path of the .exe

add \python.exe at the end

and there is still a : at the beginning.

[late cipher]

like this?

[vocal burrow]

yes

launch webui-user.bat with that

[late cipher]

YOO

I WORKS

[vocal burrow]

We ll see that when it start generating something

there is something weird with your windows setup but I don t know what exactly... it should not need us to point to the python.exe but that will do for now.

[late cipher]

yeah

[vocal burrow]

also to answer you, no need to switch to linux.

[late cipher]

yes yes

UHHH

@vocal burrow

its working

but one slight issue

its not recognizing the model type

[vocal burrow]

it s an sdxl one

[late cipher]

elaborate please

[late cipher]

i set it to XL still says the same thing

[vocal burrow]

looks good to me

[late cipher]

not generating any image tho

[vocal burrow]

what does the log says

[late cipher]

says this

[vocal burrow]

the full log not just a part of a screenshot

[late cipher]

File "C:\AI\sd-webui-forge-neo\modules\sd_models.py", line 366, in forge_model_reload
sd_model = forge_loader(state_dict, additional_state_dicts=additional_state_dicts)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "C:\AI\sd-webui-forge-neo\venv\Lib\site-packages\torch\utils_contextlib.py", line 120, in decorate_context
return func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^
File "C:\AI\sd-webui-forge-neo\backend\loader.py", line 663, in forge_loader
raise ValueError("Failed to recognize model type!")
ValueError: Failed to recognize model type!
Failed to recognize model type!
Model selected: {'checkpoint_info': {'filename': 'C:\AI\sd-webui-forge-neo\models\Stable-diffusion\ponyDiffusionV6XL_v6StartWithThisOne.safetensors', 'hash': 'e577480d'}, 'additional_modules': [], 'unet_storage_dtype': None}
Using online LoRAs in FP16: False
Model selected: {'checkpoint_info': {'filename': 'C:\AI\sd-webui-forge-neo\models\Stable-diffusion\ponyDiffusionV6XL_v6StartWithThisOne.safetensors', 'hash': 'e577480d'}, 'additional_modules': [], 'unet_storage_dtype': None}
Using online LoRAs in FP16: False
Model selected: {'checkpoint_info': {'filename': 'C:\AI\sd-webui-forge-neo\models\Stable-diffusion\ponyDiffusionV6XL_v6StartWithThisOne.safetensors', 'hash': 'e577480d'}, 'additional_modules': ['C:\AI\sd-webui-forge-neo\models\VAE\sdxl_vae.safetensors'], 'unet_storage_dtype': None}
Using online LoRAs in FP16: False
Model selected: {'checkpoint_info': {'filename': 'C:\AI\sd-webui-forge-neo\models\Stable-diffusion\ponyDiffusionV6XL_v6StartWithThisOne.safetensors', 'hash': 'e577480d'}, 'additional_modules': ['C:\AI\sd-webui-forge-neo\models\VAE\sdxl_vae.safetensors'], 'unet_storage_dtype': None}
Using online LoRAs in FP16: False

[vocal burrow]

the whole thing.

there s stuff before that

[late cipher]

i cant send the whole thign

limit to texts

i can send parts in ss

[vocal burrow]

put it in a .txt and drop file in here

[vocal burrow]

what s the size of ponyDiffusionV6XL_v6StartWithThisOne.safetensors ?

[late cipher]

6.46gb

[vocal burrow]

ah nvmd got it

OSError: The paging file is too small for this operation to complete. (os error 1455)

[late cipher]

so what do i do

[vocal burrow]

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

And set its minsize to 8gb and max size to 32

Cf for more details
https://www.thewindowsclub.com/increase-page-file-size-virtual-memory-windows

[late cipher]

@vocal burrow

this wont do much to my pc right?

[vocal burrow]

dont need it for d:

and like I said. put it to 32gb max for c:

this won t damage your pc

[late cipher]

alright?

[vocal burrow]

yes

[late cipher]

it says this

[vocal burrow]

then it s time for you to make some space on your c:

[late cipher]

alrightt doing it

alright

i restarted

im done

[vocal burrow]

then set the pagefile as explained and try again

[jagged pebble]

anyone been having similar issues to this when trying to launch stable diffusion, and is there a way to fix it? i tried looking for solutions online, but the ones i tried either didn't work, or seem way too convoluted. been getting this issue with a fresh install of python 3.10.6, git, and stable diffusion/automatic1111 installed via. cmd git clone. tried a number of webui-user.bat command lines that claim to fix the issue, but to no avail. now the only one i'm using at the moment is 'git pull'

[late cipher]

YOOOO IT WORKS NOW

THANKS

[vocal burrow]

yes that s normal automatic1111 is broken because it s pointing to some dead files. cf the pinned message for some updated guide.

[jagged pebble]

ah alright cool, many thanks man ❤️

[jagged pebble]

went ahead and followed the install guide pinned in this chat, and after reaching the part of the guide telling you to input: "pip install xformers==0.0.32" within' cmd, the following error pops up. could this be related to this other 'python.exe - Entry Point Not Found' pop-up i'm getting after trying to launch stable diffusion? stable diffusion itself seems to launch fine, but getting that specific pop-up every time i do so is a bit concerning

[ornate elk]

oh sry i updated the guide rn its has to be 0.0.32.post1

you can reload the guides site for the correct command

[jagged pebble]

ah right, thanks man ^~^ <3

[ornate elk]

no problem 🙂

let me know if it worked

[jagged pebble]

yee can confirm the new command worked (again, thanks man ^~^), but for some reason i'm still getting one of the previously mentioned errors upon trying to launch stable diffusion

[ornate elk]

can you delete boath xformers folders from the venv/lib/site-packages folder and rerun the pip install xformers command from the guide?

that should fix it

[cosmic furnace]

^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "G:\SD\SD - Forge-Webui-ZLUD\333\stable-diffusion-webui-amdgpu-forge\modules\ui.py", line 16, in <module>
from modules import sd_hijack, sd_models, script_callbacks, ui_extensions, deepbooru, extra_networks, ui_common, ui_postprocessing, progress, ui_loadsave, shared_items, ui_settings, timer, sysinfo, ui_checkpoint_merger, scripts, sd_samplers, processing, ui_extra_networks, ui_toprow, launch_utils
File "G:\SD\SD - Forge-Webui-ZLUD\333\stable-diffusion-webui-amdgpu-forge\modules\deepbooru.py", line 109, in <module>
model = DeepDanbooru()
^^^^^^^^^^^^^^
File "G:\SD\SD - Forge-Webui-ZLUD\333\stable-diffusion-webui-amdgpu-forge\modules\deepbooru.py", line 18, in init
self.load_device = memory_management.text_encoder_device()
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "G:\SD\SD - Forge-Webui-ZLUD\333\stable-diffusion-webui-amdgpu-forge\backend\memory_management.py", line 796, in text_encoder_device
if should_use_fp16(prioritize_performance=False):
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "G:\SD\SD - Forge-Webui-ZLUD\333\stable-diffusion-webui-amdgpu-forge\backend\memory_management.py", line 1102, in should_use_fp16
props = torch.cuda.get_device_properties("cuda")
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "G:\SD\SD - Forge-Webui-ZLUD\333\stable-diffusion-webui-amdgpu-forge\venv\Lib\site-packages\torch\cuda_init_.py", line 525, in get_device_properties
if device < 0 or device >= device_count():
^^^^^^^^^^
TypeError: '<' not supported between instances of 'NoneType' and 'int'

sooo any idea what can be done about this ?=\

[jagged pebble]

when i delete both the xformers folders you mentioned, then input that command into cmd, the xformer files don't seem to redownload. does "requirement already satisfied" mean that my pc thinks the files are already installed, when they're not? when i re-launch stable diffusion regardless, some cmd lines mention xformers re-installing, and after doing so, the same 'python.exe - Entry Point Not Found' error reappears

[ornate elk]

oh you have installed outside the venv. thats not good.
run this command: pip uninstall xformers -y
then run these commands one by one:
venv\scripts\activate
pip uninstall xformers -y
pip install xformers==0.0.32.post1

all in the cmd you had open inside the folder

[jagged pebble]

ay niice man, that seems to have fixed the issue! just gave stable diffusion a few test boots after inputting everything you told me to, and that one error isn't popping up anymore. xformers seems to be working as intended. you've been a huge help man, thanks a bunch! ❤️ ❤️

[ornate elk]

perfect! no problem 🙂

[ornate elk]

would need to know what you tried to to, when this error pops up

[cosmic furnace]

just launch zluda forge? it wont launch -_-

[ornate elk]

can oyu share the whole cmd log?

[gloomy bolt]

hello, the CivitAI Browser+ https://github.com/anxety-solo/sd-civitai-browser-plus.git is not updating the model previews, I think it's scanning the wrong folder or something how can I resolve this?

[ornate elk]

You still have the same error here.
ROCm agent enumerator failed: [WinError 267] Неверно задано имя папки
ROCm: version=Python311
ZLUDA support: experimental
ZLUDA load: path='G:\SD\SD - Forge-Webui-ZLUD\333\stable-diffusion-webui-amdgpu-forge.zluda' nightly=False
Failed to load ZLUDA: Could not find module 'G:\SD\Python311\bin\rocblas.dll'

Try
Uninstall HIP SDK and reinstall it at its default path not on G SD if its there

Then reboot the PC, delete the venv folder and relaunch the webui-user.bat

[cosmic furnace]

isnt it weird that neo forge with (the rock) rocm launches tho ? no issues 🤨

[cosmic furnace]

Reinstalling hip sdk didnt help at all =\

[cosmic furnace]

if forge neo that uses rocm runs fine. i guess it is fine -_- . just forge-zluda messed up somehow

[jagged crane]

who has a config for kohya_ss 4GB VRAM

[ornate elk]

Very strange, but forge neo with rocm is better anyway

[ornate elk]

You dont want to use 4GB for training, its barely possible and slow as hell
Better use Cloud Services or ask someone with more vram to train.

[misty marten]

Can someone help me with this error? I'm using comfyui-zluda

[ornate elk]

Can you show the workflow? What's your GPU?

[chilly knot]

i want duplicate space -WAnvideo on hugfaces but 50 gb limit error
i can use only CPU
Its posible to run it?