🤝｜tech-support

[night orchid]

oh no

AnimateDiff, no

[ornate elk]

that works, but reauires a lot of vram

[night orchid]

I have 12 , is that enough?

[ornate elk]

should work yes

[night orchid]

thanks, I will try it

[dusk condor]

https://private-user-images.githubusercontent.com/110560508/311155823-20f21f47-9a7b-48b5-83dd-3ace42852624.png?jwt=eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJpc3MiOiJnaXRodWIuY29tIiwiYXVkIjoicmF3LmdpdGh1YnVzZXJjb250ZW50LmNvbSIsImtleSI6ImtleTUiLCJleHAiOjE3MDk5MDI3OTEsIm5iZiI6MTcwOTkwMjQ5MSwicGF0aCI6Ii8xMTA1NjA1MDgvMzExMTU1ODIzLTIwZjIxZjQ3LTlhN2ItNDhiNS04M2RkLTNhY2U0Mjg1MjYyNC5wbmc_WC1BbXotQWxnb3JpdGhtPUFXUzQtSE1BQy1TSEEyNTYmWC1BbXotQ3JlZGVudGlhbD1BS0lBVkNPRFlMU0E1M1BRSzRaQSUyRjIwMjQwMzA4JTJGdXMtZWFzdC0xJTJGczMlMkZhd3M0X3JlcXVlc3QmWC1BbXotRGF0ZT0yMDI0MDMwOFQxMjU0NTFaJlgtQW16LUV4cGlyZXM9MzAwJlgtQW16LVNpZ25hdHVyZT02MzdiYmNmZDJlYTU3ZTliMTc5YTJlNTgxNWQzNGY4YTFjYmNkMDY4OGM1MDMxMWVlYTcyMWQyMTQ1ZGZhYTA4JlgtQW16LVNpZ25lZEhlYWRlcnM9aG9zdCZhY3Rvcl9pZD0wJmtleV9pZD0wJnJlcG9faWQ9MCJ9.FIPNI2aJQ2ELBlKLbhsDio5U2zGN9DmkSNHJLizByjg

This is the third time to tried to request to fix my bugs

And I am stuck in this issue for 1 week.

[night orchid]

@ornate elk I see that AnimateDiff is a very different extension

with mov2mov and Reactor you paste a video and a image, and the AI will try to deepfake the person in the video

[ornate elk]

ah okay

[karmic crown]

WARNING: Ignoring invalid distribution -orch (c:\users\asus\downloads\stable-diffusion-webui\venv\lib\site-packages)

where did this come from?!

i would delete venv and let a1111 make it again

[night orchid]

@ornate elk for those like me want to use mov2mov before they will update sd you can put a Git reset --hard cf2772f in the webui-user.bet

sorry bat

nah thats not working

[karmic crown]

@night orchid what error do you have? and why revert to 1.7.0?

[dusk condor]

which one?

[karmic crown]

remove the whole venv folder, it will be created again on launch

[night orchid]

@karmic crown https://github.com/Scholar01/sd-webui-mov2mov/issues/135

thats why

[karmic crown]

show me the log

i want to see exact error you are getting

it is probably the same problem as deforum had, and we fixed that

actually, the same cause, maybe not the same error

[night orchid]

@karmic crowni am realy gretfull for any help, but there are no log

[karmic crown]

?

[vocal burrow]

the console log,

[karmic crown]

cmd window

[night orchid]

nada

[karmic crown]

no errors? just nothing happens?

[night orchid]

yepp

[vocal burrow]

There has to be something

[night orchid]

SD 1.8 broke the mov2mov

[karmic crown]

if there is nothing i think i know what is the problem

[vocal burrow]

send the full log even if you don't see any errors

[night orchid]

Ok nice, i will do that

I am going out, I will try when I come back, thanks for any help today

[analog anvil]

I am trying to use img to img grpc python and the api returned the response:
answer_id: "error"
request_id: "e66e1dfd-8ca4-4ab3-b25f-6d7e7d2eb1c5:0"
received: 1709905532939
created: 1709905532951

What am I doing wrong here may I ask?

[dusk condor]

thanks it works

[vocal burrow]

Hard to say without seeing the code and/or request.
And even then I don't use api so I can't promise anything

[karmic crown]

i just checked, there is NOTHING 🙂

and reverting back to 1.7.0 with hard reset works

git reset --hard cf2772f

[vocal burrow]

yeah but there has to be a log of some sort

at least so we can check if the install has gone wrong

[karmic crown]

install is fine

it is breaking change in 1.8.0

[vocal burrow]

for it to be empty at runtime, sure ....

[karmic crown]

we fixed that for deforum

button was unclickable

here you can click button, but it does nothing

give a few minutes to try to find the buttons and to add that new one

[vocal burrow]

try to run it with '--loglevel INFO' or even DEBUG

[karmic crown]

as i said, i know exactly what is the problem

when we fix this in deforum i will add a link to the ticket what is the problem and how to solve it, but i do not want to make a PR for this, since i do not feel comfortable enough to make a PR for something i used about 2 minutes so far

but it works without a problem with the changes i just did

intresting

looks like it works on 1.7.0 too

maybe i submit PR after all

[reef smelt]

I have a problem, I get large lag spikes or like it is stalling at the end of the generation process on large images, it's been stuck at 98% for two minutes

I'm using an sdxl model

I'm also using a1111 forge

okay well it finished but it took a lot of time for some reason

[karmic crown]

https://github.com/Scholar01/sd-webui-mov2mov/pull/137

OK, I made a PR to fix mov2mov on 1.8.0 🙃

[reef smelt]

Does the size of the image have something to do with finalization of the image?

[karmic crown]

looks like mov2mov creates a panel in a different way than some other extensions, which eliminates the need for complicated changes we were testing for deforum, i sent this to deform devs to check if they can simplify their code with this

[reef smelt]

what is kohya hrfix?

[karmic crown]

please test my fix
https://github.com/Scholar01/sd-webui-mov2mov/issues/135#issuecomment-1985870161

[valid iron]

i just openned and updated Stable diffusion automatic 1111 and it made a lot of changes, now its saying this error when i try to do a basic generation "NotImplementedError: No operator found for memory_efficient_attention_forward with inputs: query : shape=(2, 4096, 8, 40) (torch.float16) key : shape=(2, 4096, 8, 40) (torch.float16) value : shape=(2, 4096, 8, 40) (torch.float16) attn_bias : <class 'NoneType'> p : 0.0 cutlassF is not supported because: xFormers wasn't build with CUDA support flshattF is not supported because: xFormers wasn't build with CUDA support tritonflashattF is not supported because: xFormers wasn't build with CUDA support triton is not available requires A100 GPU smallkF is not supported because: xFormers wasn't build with CUDA support dtype=torch.float16 (supported: {torch.float32}) max(query.shape[-1] != value.shape[-1]) > 32 unsupported embed per head: 40"

the update seemed to install and uninstall torchvision but im not good with this stuff

[vocal burrow]

what s the full log ?

[valid iron]

does this help?

some forums are saying reinstal xformers

or putting !pip install --pre -U xformers in the file edit

[vocal burrow]

the full log please , from the beginning

[vocal burrow]

yup

[valid iron]

this after restarting the command and ui

[vocal burrow]

WARNING:xformers:WARNING[XFORMERS]: xFormers can't load C++/CUDA extensions. xFormers was built for:
    PyTorch 2.0.0+cu118 with CUDA 1108 (you have 2.1.2+cu121)
    Python  3.10.10 (you have 3.10.11)
  Please reinstall xformers (see https://github.com/facebookresearch/xformers#installing-xformers)
  Memory-efficient attention, SwiGLU, sparse and more won't be available.
  Set XFORMERS_MORE_DETAILS=1 for more details
=================================================================================
You are running xformers 0.0.17.
The program is tested to work with xformers 0.0.23.post1.
To reinstall the desired version, run with commandline flag --reinstall-xformers.

Use --skip-version-check commandline argument to disable this check.
=================================================================================

[valid iron]

its been around a month since i run the wbf file

[vocal burrow]

don't ignore errors even if they re """"just at startup""""

[valid iron]

imk not verse in this stuff, so is it saying reinstall xformers

[vocal burrow]

use the --reinstall-xformers once to reinstall xformers, then remove it and see how it goes

ok so

what s in your webui-user.bat file (edit it with notepad) ?

[valid iron]

ok just copy that into the command?

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--reinstall-torch --xformers --autolaunch --api
git pull
call webui.bat

[vocal burrow]

add --reinstall-xformers to the set COMMANDLINE_ARGS line

relaunch webui-user.bat once

[valid iron]

like this?

[vocal burrow]

then re edit it and remove both --reinstall-xformers and --reinstall-torch

[valid iron]

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS--reinstall-xformers to the set COMMANDLINE_ARGS
git pull
call webui.bat

sorry ill need an example

[vocal burrow]

what

1/ edit it like this

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--reinstall-torch --reinstall-xformers --autolaunch --api 
git pull
call webui.bat

2/ run it, close it
3/ edit it like that

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--xformers --autolaunch --api 
git pull
call webui.bat

[valid iron]

thanks ill try that

i followed and it says the same error

NotImplementedError: No operator found for memory_efficient_attention_forward with inputs: query : shape=(2, 4096, 8, 40) (torch.float16) key : shape=(2, 4096, 8, 40) (torch.float16) value : shape=(2, 4096, 8, 40) (torch.float16) attn_bias : <class 'NoneType'> p : 0.0 cutlassF is not supported because: xFormers wasn't build with CUDA support flshattF is not supported because: xFormers wasn't build with CUDA support tritonflashattF is not supported because: xFormers wasn't build with CUDA support triton is not available requires A100 GPU smallkF is not supported because: xFormers wasn't build with CUDA support dtype=torch.float16 (supported: {torch.float32}) max(query.shape[-1] != value.shape[-1]) > 32 unsupported embed per head: 40

[vocal burrow]

ok what s the full log now ?

[valid iron]

i exited command ill do it again and post

not sure if the model matters

i started it all years ago and have had to fix it every update to this point and always forget to turn off auto update

[ornate elk]

you need to update your extensions,
after that delete the venv folder and relaunch

i see deforum here, and viking typing xD

[valid iron]

i dont use deforum so i can do that. so update all extensions? and sorry im not good at typing

oh the person haha my bad, all help welcome

[karmic crown]

My OCD is in the red zone when I see --autolaunch without --listen or --share 🙈

it is used to open webui only when it is not opened by defult, and that is with share or listen 🙂

[valid iron]

im new to this stuff and only know the basics of setting it up

[karmic crown]

it was nothing against you

i am just fascinated how some people who are "experts" advice new users to add usless things

[ornate elk]

yes update all the extensions in the extension tab

"Check for updates", wait until its done and close, then delete the venv folder and restart the webui-user.bat

[valid iron]

ok so hold on

[vocal burrow]

ok repeat step1 and send the full log of that

[valid iron]

so i do it right am i removing from here?

[karmic crown]

I fixed deform and mov2move to work on 1.8.0
maybe I should start accepting wishes to fix nonworking buttons on 1.8.0 on different extensions 🙂

[ornate elk]

no in the webui

there is a tab called extensions

[vocal burrow]

even before thinking about updating extensions I d say it s best to get your auto1111 to boot without any errors.

[valid iron]

oh i see, i do contrlnet and that all

[vocal burrow]

one step at a time

[ornate elk]

its faster to update and deleting the venv

will probably fix all

[karmic crown]

200% agree

[karmic crown]

this is easier and faster 🙂

[vocal burrow]

that s one way to go too :p

[valid iron]

ok just incase ill do a reboot and send logs then try the updating and ven del

after a image gen

same proccess as image

[vocal burrow]

I don t see --reinstall-xformers in there

Launching Web UI with arguments: --xformers --autolaunch --api

[valid iron]

send the batch file note edit?

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--xformers --autolaunch --api
git pull
call webui.bat

[vocal burrow]

where is --reinstall-xformers ???

told you

it was step 1

1/ edit it like this

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--reinstall-torch --reinstall-xformers --autolaunch --api 
git pull
call webui.bat

[ornate elk]

you need --xformers for --reinstall-xformers to work

[valid iron]

oh sorry i followed the steps you sent and copied the second option back in

[vocal burrow]

oh

f it

go to step 3

delete venv folder

relaunch it

update extensions

have fun

I have to go soon anyways, it will be easier this way XD

[valid iron]

its working with the edit @echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--reinstall-torch --reinstall-xformers --autolaunch --api
git pull
call webui.bat

image gen is running again, ill update my extension toojust to be safe. you guys have all been a great help, how can i thank you?

[ornate elk]

whats your gpu btw?

make sure to remove --reinstall-torch --reinstall-xformers

and add --xformers

[valid iron]

just a 2060 haha thank you so much

[ivory abyss]

AssertionError: We do not support vanilla attention in 1.13.1+cu117 anymore, as it is too expensive. Please install xformers via e.g. 'pip install xformers==0.0.16'

but i have same xformers version

[vocal burrow]

which version ? the 8 or 12gb vram version ?

[vocal burrow]

what s the full log ?

[valid iron]

i think 12

[vocal burrow]

the full log please, from the beginning

is that it ?

[daring trout]

Hi, what could I input as an addition in a prompt in order to assure that only a single animal or human appears in my final image? All the things I tried so far didn't work sadly.

[ivory abyss]

idk

is this full log?

[vocal burrow]

better

[vocal burrow]

Seen at the very beginning of the log

==============================================================================
You are running torch 1.13.1+cu117.
The program is tested to work with torch 2.1.2.
To reinstall the desired version, run with commandline flag --reinstall-torch.
Beware that this will cause a lot of large files to be downloaded, as well as
there are reports of issues with training tab on the latest version.

Use --skip-version-check commandline argument to disable this check.
==============================================================================

so yup use --reinstall-torch once to update to the proper version, then remove it

[karmic crown]

please ask that in #📝｜prompting-help this channel is for technical issues, thanks for understanding

[vocal burrow]

• "multiple humans, multiple animals" in negative prompt
• make sure you re using a resolution adapted to the model you re using (no 1024x1024 native for sd15 models)
• use controlnet to constraint the output of your subject to a certain pose
• use regional prompting to restrict your subject to one part of the image
• etc

and yes, please ask in #📝｜prompting-help for more details on that

[daring trout]

Thanks for the answer and sorry about using wrong channel 😅

[vocal burrow]

np, happens all the time

[ivory abyss]

okay i did it what can i do now

[karmic crown]

if you reinstalled those moduls try to run a1111 again

does it work? any errors?

[ivory abyss]

yes still got same error

[karmic crown]

please run pip3 list and show me what you get

[ivory abyss]

`Package Version

---

filelock 3.13.1
fsspec 2024.2.0
Jinja2 3.1.3
MarkupSafe 2.1.5
mpmath 1.3.0
mypy-extensions 1.0.0
networkx 3.2.1
numpy 1.26.4
pip 24.0
pyre-extensions 0.0.23
setuptools 69.1.1
sympy 1.12
torch 2.2.1
typing_extensions 4.10.0
typing-inspect 0.9.0
wheel 0.42.0
xformers 0.0.16`

[karmic crown]

and pip3 --version

[ivory abyss]

pip 24.0 from C:\Users\berka\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.12_qbz5n2kfra8p0\LocalCache\local-packages\Python312\site-packages\pip (python 3.12)

[karmic crown]

that is fine, as long as you have 3.10 installed as well

[vocal burrow]

Launching Web UI with arguments:
I don t see --reinstall-torch in the list of arguments

[karmic crown]

i think it would be better just to remove venv and relaunch

[vocal burrow]

you do launch sd using webui-user.bat, dont you

[karmic crown]

especially if he is launching or editing the wrong file

[vocal burrow]

it s always the easy solution yes. But I prefer to educate people when I have time for it. Plus it saves them some time as they don t have to rebuild the whole thing

[ivory abyss]

uh i launch run.bat

[vocal burrow]

run.bat ? what is that ?

screenshot your sd folder

[vocal burrow]

ok and where did you put --reinstall-torch then ?

[ivory abyss]

cmd

[karmic crown]

i guess i am forgetting that some people have much slower connection and computer than me
i am used to installing the whole a1111 from scratch including downloading of default 1.5 model in under 2 minutes 🤷🏻‍♂️

(even with purged cache) 🙈

[ivory abyss]

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

[vocal burrow]

that s not a command tho. it doesn t go in cmd

[ivory abyss]

i got this error when i start with webui-user.bat

[vocal burrow]

it s a command line arguments, it goes in webui-user.bat
at the set COMMANDLINE_ARGS= line

[ivory abyss]

do i need to do this?

[karmic crown]

- is important

[vocal burrow]

yes, and what s your gpu while you re at it ?

[vocal burrow]

oh yes you re missing one -

[karmic crown]

this is not good --reinstall-torch

[ivory abyss]

rtx 2080 ti

[vocal burrow]

--reinstall**-**torch

[ivory abyss]

okay tysm

[vocal burrow]

ok then throw --reinstall-torch --xformers in there

run it once like this

then go back to it and remove --reinstall-torch

[vocal burrow]

well it said it all,

INCOMPATIBLE PYTHON VERSION

This program is tested with 3.10.6 Python, but you have 3.12.2

[ivory abyss]

yes but i downloaded 3.10.6 version

[vocal burrow]

so uninstall python 3.12.2, install any compatible python (any python from 3.10.6 to 3.10.11), delete the venv folder sitting next to webui-user.bat

make sure to tick the "add to path'' checkbox during installation

https://cdn.discordapp.com/attachments/1002602742667280404/1075947280307331132/image-49.webp

(time for me to commute back home, see you in a bit)

[ivory abyss]

I did everything you said

but still got same error

[karmic crown]

This program is tested with 3.10.6 Python, but you have 3.12.2.

[ivory abyss]

ohh mb i have extra python in microsoft store

i'll delete it now

[old parrot]

Anyone know how to reset gradio config within python? Ran the wrong command now nothing works. Even after reinstalling

[ornate elk]

delete the config.json and the ui-config.json for a full settings reset

[old parrot]

Appreciate you as always!

[river flame]

OpenPoseXL2 worked, then I restarted A1111, it didnt work (same settings), restarted again, works again. Magic. Though the preview is not loading so I cant edit

[noble belfry]

dont worry, its ridiculously common. AI projects with use torch which is way behind in python.

[fair oxide]

I noticed when I compared my performance with other users who have the same hardware that it's very low

on 512x512 sd 1.5 i am getting max 14 it/s on 4070ti

Where I saw other users get over 30

on 1024x1024 sdxl it's under 2 it/s

Default installation of a1111 didn't mess with anything

Any ideas for what could be the reason?

[old parrot]

Have you seen this error? PermissionError: [Errno 13] Permission denied: 'C:\Windows\System32\config\BBI'

[noble belfry]

oh my god get out of system32 and stop deleting things in there

please. 😄

[young forge]

qq - Is ~30 minutes to generate a picture considered slow?

my specs

[old parrot]

Lol... I'm not THAT braindead. I'm trying to run Gradio deploy and that is the error

[vocal burrow]

1/ radeon 780m isn't a proper gpu so most likely it's only using your cpu at the moment
2/ 1920*1080 is a really high native resolution to ask for
=> yes

[young forge]

gotcha, thanks

I just read it is better to generate at a lower resolution and then upscale

[vocal burrow]

get lighting / turbo model they allow to generate at much lower steps too

[vast crest]

Hi, I noticed a problem in version 18. When I take a photo and want to save it as a preview for that model in the Lora section, it doesn't save it for me. I have to recharge the Lora every time and it appears, which has never happened with the other versions

just for informational purposes

then if there is a solution, even better 🙂

[ornate elk]

Your missing the performance args in the webui-user.bat

[fair oxide]

What do i need to type

And what do i compromise by using them

[ornate elk]

Nothing,
in the webui-user.bat you need to add --xformers --no-half-vae
To the COMMANDLINE_ARGS=

[fair oxide]

lets try

do i need quotation marks? @ornate elk

[vocal burrow]

no

[fair oxide]

oh shit

that did it

[vocal burrow]

also with --xformers will alter the randomness of your output. It won't affect the quality tho.

[fair oxide]

wtf

such a big difference

what do those two do

to be so much faster

[vocal burrow]

--xformers is a library that optimize some math stuff. --no-half-vae actually slows you down but ensure that you won't get nan error at the end.

[fair oxide]

so i am trading the ability to reproduce results

for speed?

is that what randomness means

[vocal burrow]

kinda yes.

[fair oxide]

solid

thank u very much both

[vocal burrow]

The composition should still be the same overall most of the time.

[ornate elk]

Yea and the reproducebility is like 99%

[fair oxide]

that is more than favorable

[timid coral]

Hi.

I am new to Discord.

[night orchid]

welcome

[night orchid]

I wanted to try your fix for mov2mov, but now I get errors

[paper crag]

venv "C:\SD-Zluda\stable-diffusion-webui-directml\venv\Scripts\Python.exe"
WARNING: ZLUDA works best with SD.Next. Please consider migrating to SD.Next.
fatal: No names found, cannot describe anything.
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Version: 1.8.0-RC
Commit hash: 25a3b6cbeea8a07afd5e4594afc2f1c79f41ac1a
Traceback (most recent call last):
File "C:\SD-Zluda\stable-diffusion-webui-directml\launch.py", line 48, in <module>
main()
File "C:\SD-Zluda\stable-diffusion-webui-directml\launch.py", line 39, in main
prepare_environment()
File "C:\SD-Zluda\stable-diffusion-webui-directml\modules\launch_utils.py", line 568, in prepare_environment
raise RuntimeError(
RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check
Press any key to continue . . .

i heard im not the only one that got this error

i installed the Zluda version yesterday

was working very well

now i get this bs

[karmic crown]

than you probably did something wrong, two other guys confirmed that it works properly for them too

show me the log

[night orchid]

my bad I explaination is terrible, what I wanted to say was that I cant try the fix, because of an error

[karmic crown]

download the archive from attachment, extract zip, find the file on your computer and replace it
restart webui and it will work
but first update both a1111 and extension

if it still not work, show me the log from cmd

[edgy blaze]

Hello. I'm interested in accessing Stable Diffusion and downloading the latest version of the model. However, I'm finding it overwhelming with so many links and download options available. I'm unsure where to find the most straightforward version of Stable Diffusion. Are there any instructional videos that could help me navigate this situation? If not, are there any guides or individuals who can provide assistance?

[night orchid]

https://pastebin.com/bGyAxxPx

[onyx solar]

Literally #1080946152318443610.

[karmic crown]

your problem is onnx, not a1111 or mov2mov

[night orchid]

ok but.... i did not have that error before

[karmic crown]

even if you delete mov2move from your extensions compleatly, your webui will not work

you need to fix your webui first

[night orchid]

yes that much I understand 🙂

[paper crag]

did you manage to fix this error?

[edgy blaze]

Is this what I'm meant to be looking for?

[karmic crown]

check pined messages in this channel and you will find two amd guides there, follow one of them

[sweet basin]

it fixed by itself 1h later

you getting it?

[paper crag]

yeah, same error

installed zluda yesterday, everything was well, opened my pc now, im getitng this error

[night orchid]

search youtube for stable diffusion automatic 1111

[onyx solar]

Stable Diffusion is not the software, it's the image inference model. Look for Automatic1111.

[edgy blaze]

Noted, thank you.

[night orchid]

@onyx solar thats correct, but the automatic1111 webui will install stable diffusion 1.8

[onyx solar]

Hate to do this, but ACKSHUALLY, it's Automatic1111 that is version 1.8; Stable Diffusion's models that work within A1111 are not.

[sweet basin]

yeah I had same problem. was working fine but next morning I got this error

but tried later and it worked again

[night orchid]

@onyx solar happy to have experience peoples like you

[karmic crown]

just add --skip-torch-cuda-test to COMMANDLINE_ARGS

we use that on macs since day one

[sweet basin]

working fine right now

[onyx solar]

I do really hate to pull the "actually" card, though. 🙂

[night orchid]

absolut no problem

[vocal burrow]

it's important though.... too many people confuse the two.

[night orchid]

I ahppy to be corrected

[onyx solar]

Eventually, this won't be as much of an issue.

[night orchid]

hope so

but for now 1.8 is Automatic1111 webui, and stable diffusion 2.x , right?

[onyx solar]

I believe that A1111 supports multiple model types, including 1.5, 2.X, XL, & Cascade.

[night orchid]

You are right

[onyx solar]

Personally, I prefer ComfyUI, but a lot of people find node-based stuff has a bigger learning curve.

[fair oxide]

question, im training my first lora, and im at this part of the cmd, is it still in a process, i cannot tell

[vocal burrow]

nope, it crashed

[fair oxide]

ah, got a solution?

[night orchid]

@onyx solar how about SD 3.0 , that one will also be free right? and supported by Automatic1111

[onyx solar]

Based oon what emad seems to be saying, I'd imagine it will.

[vocal burrow]

no clue sorry, I don't do training :/

[onyx solar]

Eventually.

[fair oxide]

i see

im smart, it was the model

cause pickletensors is like, a model thing

[high oar]

Hey, i'm trying to install Automatic1111 on Linux Mint, but because of a weird qwerk in my GPU i can only use this (pictured) driver, and it keeps downloading an incompatible version of torch. Same issue occurs with the text-generation web ui

Any help would be greatly appreciated!

[karmic crown]

the driver version and torch version are not directly connected

show me the gide guide that you are following

and what you did

[fair oxide]

does everything look according now, i hope

if anyone is good with loras, lmk

[reef smelt]

Is there an extension that does autotag completion but like linting? I find myself getting lost in my own prompts or others prompts because I can't find the keyword that does something

[reef smelt]

I know somethings

but what do you mean?

[fair oxide]

it just crashed at 30 percent, had to restart the training

ill show

[reef smelt]

oh.. training

[fair oxide]

yeah im training a lora

[reef smelt]

then I'd say I don't know a lot

[fair oxide]

yeah its my first time trying learning as i go

here is the crash in cmd

it got to about 1800 steps

[night orchid]

@karmic crown now ! no errors, so now I can test your solution

[fair oxide]

im so stupid i was doing a dreambooth, not a model

lmfao

like not a lora, but a dreambooth "model"

i was in the wrong kohya_ss tab

[night orchid]

@karmic crown time to test (drumrolls)

I can confirm that it is indeed working, thanks man, well done !

[karmic crown]

please comment on github and say that it works for you too, so it will be merged faster

[night orchid]

indeed

done !

[karmic crown]

they should be grateful to you, i would not even look at it if you hadn't complained here 🙂

i do not use that extension

[fossil mortar]

hi guys i encountered problem how i fix it i press any key but its didnt happen anything, thank you for now
RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check Press any key to continue . . .

[karmic crown]

try to add --skip-torch-cuda-test to COMMANDLINE_ARGS, as we are doing on Macs

for example, those are mine on Mac

COMMANDLINE_ARGS="--skip-torch-cuda-test --opt-sub-quad-attention --upcast-sampling --no-half-vae --medvram-sdxl --use-cpu interrogate --no-download-sd-model"

[fossil mortar]

" '--skip-torch-cuda-test' is not recognized as an internal or external command,
operable program or batch file. "

[karmic crown]

show me your complete command line args

[fossil mortar]

i send in dm

Do you mind if I send it here?

[karmic crown]

please send here next time

This program is tested with 3.10.6 Python, but you have 3.11.7.

this is your first problem

you need to fix this first

you cant use 3.11

you need to use 3.10.x

[fossil mortar]

oh sorry 😕

[karmic crown]

this warning Use --skip-python-version-check to suppress this warning. might be just a consequence of using wrong python version and not a real message that you should see

but we will see what happens when you fix python version

[fossil mortar]

Should I start all the process from scratch?

[karmic crown]

nope

wait

try this

open cmd and type

python3.12 --version
python3.11 --version
python3.10 --version
python3 --version
python --version

and show me the result

sorry all 5

just to be sure what you have

[karmic crown]

press enter on that last one too

[fossil mortar]

python not found

[karmic crown]

Python bulunamad - what this means?

[fossil mortar]

python not found

[karmic crown]

Are you getting the same message when you run the Python --version ?

have you uninstalled 3.11 just before you run that? if you did it is fine

[fossil mortar]

yeah

[karmic crown]

ok, you have none

install 3.10.6

[fossil mortar]

Yes, I deleted it when you said wrong version.

[karmic crown]

find a link in a guide in pinned message

Preparation:
You need to install Git 64bit: https://git-scm.com/download/win
and Python 3.10.11, 64bit: (any python above or 3.10.6) Not 3.11, 3.12 https://www.python.org/downloads/release/python-31011/
And check "add python to path" when installing Python.

this

you do not need to install git again

but install correct python version

[fossil mortar]

I downloaded

[high oar]

The Automatic1111 GIThub install guide and a Google search of "which python torch version is compatible with NVIDIA Ubuntu driver 470"

[fossil mortar]

Why does it still say Python not found?

[karmic crown]

you need to close the cmd and reopen it

it should detect it than

[karmic crown]

please show me the result when you run this:

python3.12 --version
python3.11 --version
python3.10 --version
python3 --version
python --version

[high oar]

Ok, I'm not at my computer rn but I will in an hour or so

[fossil mortar]

Microsoft Windows [Version 10.0.22621.3155]
(c) Microsoft Corporation. All rights reserved.

C:\Users\mevlu>python3.12 --version
'python3.12' is not recognized as an internal or external command,
operable program or batch file.

C:\Users\mevlu>python3.11 --version
'python3.11' is not recognized as an internal or external command,
operable program or batch file.

C:\Users\mevlu>python3.10 --version
'python3.10' is not recognized as an internal or external command,
operable program or batch file.

C:\Users\mevlu>python3 --version
Python bulunamad
C:\Users\mevlu>python --version
Python bulunamad
C:\Users\mevlu>
Where am I making mistakes?

[karmic crown]

you need to have python 3.10.x, so if you have 3.11 or 3.12 there, you might need to change python verson (or path) in webui-user.sh
we will see

[fossil mortar]

i closed but still same

[karmic crown]

add to path probably

[ornate elk]

open up a cmd and type
where python

and then run
python --version

[fossil mortar]

C:\Users\mevlu>python --version
Python 3.10.11
🙂

[karmic crown]

CS1o, when i see the next Mac user I will let you suffer like i just did with his Windows install 🤣

[karmic crown]

remove venv remove inside stable-diffusion-webui, and run webui again

it will create new venv for you

[upbeat crane]

omg, its the 'what's in your webui-user.bat file' and the 'delete your venv' dudes at it again

hehe teasing, cheers and happy friday ya filthy animals

[ornate elk]

hey whats in your webui-user.bat?

😛

[upbeat crane]

i'll show you mine if you show me yours, mister

[karmic crown]

i do not care about webui-user.bat, show me your webui-user.sh 🙂

[upbeat crane]

does mac work with that file, like linux ppl do

[karmic crown]

macOS was initially based on FreeBSD, so it is Unix-like (same as Linux)

[upbeat crane]

DM'd u earlier btw

few min ago its nothin important just talkin about deforum shit

things started coming back around again last night after you went to bed thanks to a couple specific youtubers

[karmic crown]

i am trying to avoid using dm, to be honest

[upbeat crane]

homies with no ego in the game and spreading knowledge

whys that lol

[karmic crown]

i am usually extremely busy, so if i try to chat with people in dm, i will not do anything else

nothing personal

[upbeat crane]

mmkay it wasnt seeking troubleshooting but as you wish

[upbeat crane]

just kinda wrapping up a few thoughts from our long ass convo

[karmic crown]

as i said, nothing personal 🙂

[upbeat crane]

and i didnt wanna spam a channel about stuff nobody else but we were interested in

i get it, im a server admin of a pretty good size server

[fossil mortar]

''' (Automatic1111_olive) C:\Users\mevlu\stable-diffusion-webui-directml>env\Scripts\activate
The system cannot find the path specified.

(Automatic1111_olive) C:\Users\mevlu\stable-diffusion-webui-directml>pip install httpx==0.24.1
Collecting httpx==0.24.1
Downloading httpx-0.24.1-py3-none-any.whl (75 kB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 75.4/75.4 kB 1.4 MB/s eta 0:00:00
Collecting httpcore<0.18.0,>=0.15.0
Downloading httpcore-0.17.3-py3-none-any.whl (74 kB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 74.5/74.5 kB ? eta 0:00:00
Collecting sniffio
Downloading sniffio-1.3.1-py3-none-any.whl (10 kB)
Collecting idna
Using cached idna-3.6-py3-none-any.whl (61 kB)
Collecting certifi
Using cached certifi-2024.2.2-py3-none-any.whl (163 kB)
Collecting h11<0.15,>=0.13
Downloading h11-0.14.0-py3-none-any.whl (58 kB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 58.3/58.3 kB 153.9 kB/s eta 0:00:00
Collecting anyio<5.0,>=3.0
Downloading anyio-4.3.0-py3-none-any.whl (85 kB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 85.6/85.6 kB 1.2 MB/s eta 0:00:00
Collecting exceptiongroup>=1.0.2
Downloading exceptiongroup-1.2.0-py3-none-any.whl (16 kB)
Collecting typing-extensions>=4.1
Using cached typing_extensions-4.10.0-py3-none-any.whl (33 kB)
Installing collected packages: typing-extensions, sniffio, idna, h11, exceptiongroup, certifi, anyio, httpcore, httpx
Successfully installed anyio-4.3.0 certifi-2024.2.2 exceptiongroup-1.2.0 h11-0.14.0 httpcore-0.17.3 httpx-0.24.1 idna-3.6 sniffio-1.3.1 typing-extensions-4.10.0

[notice] A new release of pip is available: 23.0.1 -> 24.0
[notice] To update, run: python.exe -m pip install --upgrade pip

(Automatic1111_olive) C:\Users\mevlu\stable-diffusion-webui-directml> '''

[karmic crown]

(Automatic1111_olive) C:\Users\mevlu\stable-diffusion-webui-directml>env\Scripts\activate
The system cannot find the path specified.

looks like you activate env, and then run the script from there

and scripts trying activate the same venv

and if i remember correctly olive make some problems

CS1o knows the most about AMD

[ornate elk]

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

[upbeat crane]

nice

[fossil mortar]

rx 6800 XT

Where do I look webui-user.bat

[upbeat crane]

ur main stable diffusion folder just open it and post him the contents here

stable-diffusion-webui folder

open in notepad or somethin

[ornate elk]

did you do a reinstall or did you wanted to fix a broken old webui?

[ornate elk]

im asking because there is a better webui version now for AMD

[fossil mortar]

reinstall

[ornate elk]

okay then follow this guide please, the webui will be much faster and you wont run out of vram with a 6800xt:
#🤝｜tech-support message

in Ui fact its the same as directml webui but with the new zluda backend

[fossil mortar]

okay, thank you all for all things

[ornate elk]

no problem, if you have any questions at the install steps, feel free to ask

[fossil mortar]

thank you again 🙂

[hushed dirge]

Hello, please could someone guide me, I am using stable diffusion on google colab and I installed the deforum extension but the deforum tab does not appear. Please any help will be appreciated

[upbeat crane]

are you in the deforum discord?

deforum tab started dissapearing from auto1111 when 1.8 released

i run everything (including deforum) but locally, so im a bit inadequte with collab related issues

i would assume that your notebooks pulling the 1.8?

basically, right now, everything should be back on track, so i would triple check you have the latest version of deforum

so that if you are on 1.8 it should reappear

but i recommend joining the deforum discord and then using the search box or just asking in the #collab-notebook channel

maybe someone here has better info not sure

[hushed dirge]

Thanks very much , will do so

[upbeat crane]

right on

[hushed dirge]

Please do you have the link to the discord channel

[upbeat crane]

yeah

if it doesnt let me post it here i will dm you an invite

DM'd

[hushed dirge]

Thanks, in it now

[upbeat crane]

check the PIN in the collab channel first

before pinging them

[eternal umbra]

Hy, I'm on my phone now, and the prompt does not appear when I type ” /imagine ” any idea why?

[onyx solar]

#1047610792226340935

[eternal umbra]

oh, ok, I thought I'm doing something wrong 😅, first time on phone 😁

thanks and sorry for disturbing

[onyx solar]

You had the decency to ask questions and not just send prompts into various channels. That's worth answering and not a bother.

[upbeat crane]

what does "on the phone" mean?

ohh like he is using his phone to try and interact with the bot haha, i thought he was sayin like "hey im on the phone here, so yall hurry up and figure this out for me" funny

[eternal umbra]

usually I'm using discord from PC, but now I'm using it from phone app

[upbeat crane]

that cracked me up

he didn't mean "im in the check-out line at walmart, can you nerds bang out this issue for me im busy"

hehe

sorry

[onyx solar]

I think the key difference was that he said "on my phone" not "on the phone". 😄

[fossil mortar]

venv "C:\SD-Zluda\stable-diffusion-webui-directml\venv\Scripts\Python.exe"
WARNING: ZLUDA works best with SD.Next. Please consider migrating to SD.Next.
fatal: No names found, cannot describe anything.
Python 3.10.11 (tags/v3.10.11:7d4cc5a, Apr  5 2023, 00:38:17) [MSC v.1929 64 bit (AMD64)]
Version: 1.8.0-RC
Commit hash: 25a3b6cbeea8a07afd5e4594afc2f1c79f41ac1a
Traceback (most recent call last):
  File "C:\SD-Zluda\stable-diffusion-webui-directml\launch.py", line 48, in <module>
    main()
  File "C:\SD-Zluda\stable-diffusion-webui-directml\launch.py", line 39, in main
    prepare_environment()
  File "C:\SD-Zluda\stable-diffusion-webui-directml\modules\launch_utils.py", line 568, in prepare_environment
    raise RuntimeError(
RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check
Press any key to continue . . .

[ornate elk]

which step?

[fossil mortar]

last launch webui-user.bat

[ornate elk]

okay then show me your paths in the environment path

[ornate elk]

you need to change the rocm path to
%HIP_PATH%bin

[high oar]

    python3.11 --version
    python3.10 --version
    python3 --version
    python --version
Command 'python3.12' not found, did you mean:
  command 'python3.10' from deb python3.10 (3.10.12-1~22.04.2)
  command 'python3.11' from deb python3.11 (3.11.0~rc1-1~22.04)
Try: sudo apt install <deb name>
Command 'python3.11' not found, but can be installed with:
sudo apt install python3.11
fish:
python3.11 --version
^
Python 3.10.12
Python 3.10.12
Command 'python' not found, did you mean:
  command 'python3' from deb python3
  command 'python' from deb python-is-python3
fish:
python --version
^```

[ornate elk]

hey, what was your inital issue?

[karmic crown]

both of those returns the same version?

python3.10 --version
python3 --version

3.10.12?

[high oar]

@ornate elk ^^^

[high oar]

yup

[edgy blaze]

Good day. I'm currently in the middle of installing AUTOMATIC1111. However, during my installment. When I went into the Webui-user folder for my windows batch file, I ran into some trouble I hadn't expected. Is this something that is meant to happen and something that may potentially break something? Or should I simply carry on.

[karmic crown]

ok, that is fine
show me your webui-user.sh now

[ornate elk]

hey, whats your gpu?

[edgy blaze]

One second please

[karmic crown]

ok, i will need log too, but paste the actual file here

[high oar]

Which log?

[edgy blaze]

@ornate elk AMD Radeon RX 7600S, AMD Radeon (TM) Graphics

[karmic crown]

i asked for an actual webui.sh file, but you posted the log when you run it

[high oar]

Ohh, i see. Apologies!

[ornate elk]

then you tried to install the wrong webui.
AMD gpus need a different one.

to start clean:
Open up a cmd and run
pip cache purge

then delete your whole stable-diffusion-webui folder.

And then follow my AMD guide from here:
#🤝｜tech-support message

[karmic crown]

sorry, webui-user.sh
i am sleepy 1am here

[edgy blaze]

Good to know. Thank you very much.

[high oar]

np

I tried to manualy add the torch version, normally EVERYTHING is commented out

[ornate elk]

np, if you have any questions feel free to ask

[karmic crown]

replace

export TORCH_COMMAND="pip install torch --index-url https://download.pytorch.org/whl/cu118"

with

export TORCH_COMMAND="pip install torch==2.2.0 torchvision==0.17.0 torchaudio==2.2.0 --index-url https://download.pytorch.org/whl/cu118"

remove venv folder
and run ./webui.sh again

if that does not work, try previous version, with this line

export TORCH_COMMAND="pip install torch==2.1.2 torchvision==0.16.2 torchaudio==2.1.2 --index-url https://download.pytorch.org/whl/cu118"

if 2.2.0 can work, use it
if not use 2.1.2

[high oar]

Ok, thank you so much. If I have any further issues I'll let you know!

[edgy blaze]

Hello again! I have another question, if you don't mind. I'm completely new to all of this. I just started installing this software a few hours ago. Where can I find the "1.5 based model" from my other existing DirectML web UI? Right now, this is my only DirectML, so should I ignore this step? @ornate elk

[high oar]

@karmic crown I got it working but it's not doing anything when I click generate

[karmic crown]

show me the log now

[ornate elk]

oh yes, then you can ignore it. it will get downloaded at the install

[edgy blaze]

Thank you. And, There seems to be an issue. Though I've been following your steps. I still ran into a road block letting me not be able to download it properly. Do you have any advice as to how I could fix this?

[old parrot]

Do 1.0 SD Loras work with stable Forge?

[karmic crown]

what dpkg -l |grep malloc shows?

[ornate elk]

have you moved the files already?
the two .dll files from zluda, into torch/lib

[torpid lantern]

Hello, beginner here. Does anyone know why I get this error? This is what I get after attempting to train a lora.

[high oar]

no output

[edgy blaze]

Let me check. Perhaps I haven't done that.

[karmic crown]

sudo apt-get install libgoogle-perftools4 libtcmalloc-minimal4 -y

this is for ubuntu, but lets hope it would work there too

or apt install, whatever you have there

i mostly use ubuntu on servers

if that does not help, check https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/10117

[high oar]

Still nothing

i tried to disable TCmalloc altogether too

[karmic crown]

what was the output of this?
sudo apt-get install libgoogle-perftools4 libtcmalloc-minimal4 -y

[high oar]

[sudo] password for jonah:
Reading package lists... Done
Building dependency tree... Done
Reading state information... Done
The following NEW packages will be installed:
  libgoogle-perftools4 libtcmalloc-minimal4
0 upgraded, 2 newly installed, 0 to remove and 16 not upgraded.
Need to get 311 kB of archives.
After this operation, 1,433 kB of additional disk space will be used.
Get:1 http://archive.ubuntu.com/ubuntu jammy/main amd64 libtcmalloc-minimal4 amd64 2.9.1-0ubuntu3 [98.2 kB]
Get:2 http://archive.ubuntu.com/ubuntu jammy/main amd64 libgoogle-perftools4 amd64 2.9.1-0ubuntu3 [212 kB]
Fetched 311 kB in 9s (35.5 kB/s)
Selecting previously unselected package libtcmalloc-minimal4:amd64.
(Reading database ... 571067 files and directories currently installed.)
Preparing to unpack .../libtcmalloc-minimal4_2.9.1-0ubuntu3_amd64.deb ...
Unpacking libtcmalloc-minimal4:amd64 (2.9.1-0ubuntu3) ...
Selecting previously unselected package libgoogle-perftools4:amd64.
Preparing to unpack .../libgoogle-perftools4_2.9.1-0ubuntu3_amd64.deb ...
Unpacking libgoogle-perftools4:amd64 (2.9.1-0ubuntu3) ...
Setting up libtcmalloc-minimal4:amd64 (2.9.1-0ubuntu3) ...
Setting up libgoogle-perftools4:amd64 (2.9.1-0ubuntu3) ...
Processing triggers for libc-bin (2.35-0ubuntu3.6) ...```

[karmic crown]

ok

it is nvidia, right?

[high oar]

yup

[karmic crown]

open webui-user.sh

and find
#export COMMANDLINE_ARGS=""

change that to
export COMMANDLINE_ARGS="--xformers"

do not forget to uncomment it

[high oar]

    PyTorch 2.1.2+cu121 with CUDA 1201 (you have 2.2.0+cu118)
    Python  3.10.13 (you have 3.10.12)
  Please reinstall xformers (see https://github.com/facebookresearch/xformers#installing-xformers)```

[karmic crown]

python shold be fine
but try to change torch
i sent you what you should use for 2.1.2 above

[edgy blaze]

Is this the right folder I should use inside my Path? Or am I meant to add this one@ornate elk

[karmic crown]

when you change that line, remove venv folder and launch it again
a1111 will create a new venv with torch 2.1.2

i need to go, 2am here
i hope you will make it work

[high oar]

Oh, of course. Thank you!

[karmic crown]

if not, ask here, someone will hopefully help

if not i will check tomorrow

good night

[high oar]

gn

Oh wait, ik the problem

none of the buttons are doing anything

@ornate elk Would you perhaps know anything?

[ornate elk]

you need to rename the folder of the second screenshot to Zluda

[edgy blaze]

Alright

[edgy blaze]

Sorry again in the case I seem rather moronic. I get relatively lost in these things. So am I meant to use the Folder in the path or am I meant to use the AMD64 in the path?

I'm inclined to believe I'm meant to use the AM64 since I put it in the C:Drive like the steps wrote.

[ornate elk]

i dont get the question

[high oar]

guys help I tried to import a different model for craps and giggles and now my whole computer's frozen!!!

[edgy blaze]

Forgive me, worded that a little strange. When I go into 'Edit Environmental Variable' and click the browse tool. There are two things that I am able to select from. The 'Zluda Folder' that I created on step 2. And the Zluda Zip File I downloaded and moved to the Zip File. Which am I meant to use inside my Edit Environmental Variables?

If I worded that correctly

[torpid lantern]

Just ran my first lora. Is it normal for the steps at the bottom of the screenshot to remain at 0% 1/45 epoch with no time elapsed for 10 minutes? my 3070 is at 98% and ram is at 50

I'm guessing batch size is a bit too crazy since it's set to 8. Will lower that

[ornate elk]

so to clarify that:
you have a SD-Zluda Folder where your directml webui is located in. (You dont need to add that to your paths)

Then you have the .zip file called ZLUDA-windows-amd64
you unzip it as folder to the C: drive and rename the folder for it to
Zluda
this folder path has to be added as environment path
then you add a second path where you just paste this in: %HIP_PATH%bin

[edgy blaze]

Ah. That clarifies it better. Thank you once again.

[ornate elk]

no problem 🙂 maybe i should update that to the guide

[edgy blaze]

Perhaps. I'm a person who is easily confused. So it's really a matter of choice.

[ornate elk]

no worries, my english wording is sometimes not good ^^

so if i can improve it for everyone to understand i will do

[humble cobalt]

hello, just asking a quick question but i was curious how does people use models on SDXL if SDXL is a model ?

[ornate elk]

hey, we dont use models "on" SDXL, we use SDXL models on any webui that supports them.
like automatic1111, comfyui, fooocus,

people often get confused and think sdxl is its own programm or does need one. Thats not the case

[humble cobalt]

because i tried using Animagine XL v3 but it doesnt seem like it want to have a good quality and when i search PNG info i always see some "SDXL" stuff but i dont really know how does it works

[edgy blaze]

I did everything I had to do and moved cublas64_11.dll and cusparse64_11.dll and overwrote them. Though Torch is still not able to use the GPU. @ornate elk

[ornate elk]

SDXL models are trained on a 1024x1024 resolution
make sure you are near that

for good quality

[ornate elk]

you did the first step wrong xD 🥲

[humble cobalt]

does it still use SDXL if i didnt download the model itself ? the base and refiner

[edgy blaze]

Care to subtly elaborate?

[ornate elk]

no, if you dont load the sdxl model or refiner in, it wont be used

also dont use refiner

its not that good

[ornate elk]

"Make a new Folder on your Drive (not on Desktop, Downloads, Documents, Programms, Onedrive) and name it SD-Zluda."

and yours is here:

[humble cobalt]

okay okay thanks :))

[ornate elk]

also for sdxl you need the sdxl vae

[edgy blaze]

Goodness. OneDrive is really messing me over here. Is there a command prompt I can use to quickly move that into my drive?

Or do I just have to click and drag it to my C:Drive

[ornate elk]

nope, you need to git clone again and pls dont name it Zluda

name it SD-Zluda

so you and i wont get confused later

[humble cobalt]

idk, im trying to reproduce the example there are on animagine XL so i can really see if i produce the same quality

but the hires. fix takes a lot of time lol

[edgy blaze]

Ah so I have to start from scratch?

[ornate elk]

yes

[edgy blaze]

Gotcha

[ornate elk]

but you can leave hip installed

and python etc

[edgy blaze]

Alright

[ornate elk]

whats your gpu btw?
you may need performance adjustements for sdxl (in your webui-user.bat)

[humble cobalt]

3060 Ti

i know 8Go VRAM isnt enough lol

[ornate elk]

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

it is enough (but upscale is hard)

[humble cobalt]

not much, only xformers and autolaunch

[ornate elk]

okay then you need to add --medvram-sdxl

that will boost sdxl models

and make sure to use the sdxl-fp16 vae

[humble cobalt]

does it work if i put --medvram ?

so it work for any model

[ornate elk]

but that would slowdown your 1.5 models, for no reason

--medvram-sdxl applies an optmization for the large sdxl models, no need for 1.5

[humble cobalt]

oh okay

and what does medvram do ?

[ornate elk]

it does the same like --medvram-sdxl but for everything,
in fact --medvram splits the models into smaller packages so its easier to load them into the GPU.
thats good for sdxl and loads them faster, but bad for 1.5 as its not necessary (because they are smal)

[humble cobalt]

okayyy thanks ! Didn't know that, before, i was thinking it was just reducing quality lol

[ornate elk]

nope, there is no command to reduce the quality

[humble cobalt]

the vae you was talking about is "diffusion_pytorch_model.fp16.safetensors" ?

the only one with fp16 i saw in vae folder of sdxl base

[ornate elk]

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

[humble cobalt]

okay thanks a lot

[ornate elk]

no problem

[ornate elk]

also if you upscale, set the hires steps to 10

that should be faster and is enough for a good quality

@edgy blaze hows it going?

[edgy blaze]

Currently waiting for this to install Torch so I can copy and paste cublas64_11.dll and cusparse64_11.dll into it

[ornate elk]

how does your folders and paths look?

you should have a SD-Zluda folder (where the webui is in)
and a C:/Zluda Folder where the dlls are in

[ornate elk]

Perfect

[edgy blaze]

Okay I've pasted cublas64_11.dll and cusparse64_11.dll into lib like the steps instructed. Should I purge these that are still in the other place?

[ornate elk]

You can leave them in there

[edgy blaze]

Alright

To my surprise this time. Still did not work. @ornate elk

[ornate elk]

Is your AMD driver on 24.1.1 or higher?

[edgy blaze]

Let me go and check.

[edgy blaze]

Unfortunately I am not on 24.1.1.

Suppose I have to update that somewhere.

[ornate elk]

Yes you need to update it

Download it from the official AMD site

24.2.1 is the latest Adrenalin Edition

After the Update, restart the PC

I'm off now gn

[edgy blaze]

Good night. Thank you for the help.

[torpid lantern]

Hey, can I get some suggestions on what I can do to lower this hoorendous dreambooth eta? running with a 3070 (8 VM)

[inland zephyr]

Is it possible to make an old SD1.5 checkpoint usable as an XL model, or does it have to be trained from the beginning all over again?

[deft hemlock]

mods did u answer a person who had problems especially peformance with the new automatic1111 version? ill follow their steps so that u dont have to bother helping me

https://tenor.com/view/холодильник-speech-bubble-gif-25663959

https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/15143

this github basically summarizes most of my problems

[scarlet hill]

running an rx 7800xt, just downloaded sd.next and running it with zluda and directml, should the first generation take ages to run?

seems to only happen with hires, gets stuck on 2% and doesn't shift

[formal trout]

with an RX 6700 I can only hi-res fix to 1.2 a 1024x1024, or I get out of memory in Automatic. Not expert about that but that what it works. Maybe test lowers values?

[scarlet hill]

Got it working, looks like it was just a memory issue

Is there any hope for AMD GPUs getting better support going forward with SD?

[fossil mortar]

it didnt work

[fair oxide]

not sure if you are supposed to use directml and zluda at the same time. i just start it with --use-zluda and it works 🤷🏻‍♂️

[kind wyvern]

Hi, I would like to use dark mode on 1111 (v1.8.0), but nothing changes even if I specify it in webui-user.bat.
Am I missing something else to do? My environment is running using MacOS 12.7.2, Stability Matrix v2.9.0.

[night orchid]

set it in your browser

thats what I did, I use Win10 and Google Chrome

https://www.alphr.com/safari-dark-mode/

[hoary plaza]

thoughts on StabilityMatrix?

[vocal burrow]

command line args start with two `-``, not one

[kind wyvern]

thank you. I use Chrome and the 1111 remains bright even when I set its settings to dark mode 🥲

[kind wyvern]

Thank you. I increased it to 2, saved, restarted chrome and 1111, But still the theme remains bright 🤔

[vocal burrow]

you're on mac. so you're not using the webui-user.bat file

those are for windows users.

you should be modifying webui-user.sh and launch sd using webui.sh

there will be a similar lines in there just commented with a # before

Check out the pinned messages btwn there's a script for mac user in there that will help you get the most out of your machine.

[idle burrow]

Hey Guys,

I've been using Dreambooth on runpod quite a bit last year (about 12 months ago) and I'm just getting back into it.
Obviously some things have changed and I find myself getting quite frustrated with the changes they made to the notebook.

It used to be that you just create a folder named "training_samples" and put your images in there.
Then they changed it to the imgur method which seemed more complicated but also worked.

Now they have a new cell that just spits out a bunch of errors.

One of the errors tells you to just put the images into the folder that the cell created, which doesn't work because after doing that when i run the training i get an error: images/folder not found
If I don't run the "error cell" at all and manually create the folder "training_images" and put the images there I get the same or similar error when running the training cell

Has anyone here recently used the notebook and can explain how you made it work?

I would also like to know:

Can I use Dreambooth to train a SDXL model or does it only work with 1.5 models?
Can I use a model from civitAI that was trained on 1.5? (or in general does Dreambooth work with safetensors or only ckpt? )

Thank You

[errant wren]

any local sdxl with comfyui guides for me to learn? just chug me a document link and i can start reading

[fossil mortar]

I couldn't do this either

[kind wyvern]

Thanks a lot! This seems like pretty valuable information, but I use Stability Matrix to install and boot A1111 / Comfy-UI.
Is it okay to use "version 0.0.6 of a1111-setup script" even in that environment?

[vocal burrow]

maybe not then

but you'll sill want to use webui-user.sh to set things up and run webui.sh

[queen plank]

Error occurred when executing VHS_VideoCombine:

Could not find ffmpeg

File "D:\ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\ComfyUI\execution.py", line 152, in recursive_execute
output_data, output_ui = get_output_data(obj, input_data_all)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

Could someone help me here?

Wait

I think I fiexed it

[ornate elk]

Why?

Show me your environment paths and the two folders

[hoary plaza]

anyone use stablility matrix?

[manic snow]

yup!

[fossil mortar]

Did it happen?

[hoary plaza]

ooo how good is it?

[manic snow]

it's very convenient mostly. it has a model browser, it automatically creates symlinks so that you don't have to copy all your models and it's very easy to install UIs

[noble belfry]

That is a nice feature. for windows, I'd love with a1111,comfyui and the rest all just used a single model folder we had an environment path. i mean my models folder is far bigger than my music, my videos, my pictures folders and local AI seems to be going that way.

i know we all can create symlinks, but it would cut down on having to teach it/learn how/manually change for users. i'd be interested in the % of users that have more than one client at this point. client/studio/UI/frontend whatever we're calling them. still not sure.

[vestal laurel]

pls help

i've been trying so hard to find a solution i can't

[thorny iron]

So I did 150 sampling steps

It has been 99% for around 5 mins now

Is that stuck or..

[vestal laurel]

at least you can generate the pic, mine won't even work idk why

it says torch not compiled idk what that is

[thorny iron]

Erm..

https://www.datacamp.com/tutorial/how-to-run-stable-diffusion

I think I followed this guide to install it

https://www.youtube.com/watch?v=TLbVrabNuHY&t

Maybe watch this

You missed a step during installation it looks.

Wew 150 steps took 15 mins

[vestal laurel]

@thorny ironytsm

tysm*

[vocal burrow]

would need to see the full console log and your hardware specs

[fossil mortar]

@ornate elkthank you for all i did it

[vocal burrow]

why 150 steps, that's so high for no reasons.
would need to see full log and screenshot of your txt2img webpage

[ornate elk]

Ah perfect np 🙂

[vestal laurel]

alright wait

[fossil mortar]

and the last thing can the interface be made dark theme?

[vestal laurel]

i'm sending the console log

[vocal burrow]

it's missing the central piece .... the gpu

[thorny iron]

For fun, it was interesting, and a very clear image honestly. ';d

[vocal burrow]

Still it makes no sense to go this high. You passed the point of no return, got all around the earth and passed it again at mach 3.

let me illustrate

[vestal laurel]

here

[vocal burrow]

https://cdn.discordapp.com/attachments/1002602742667280404/1120475781349527553/xyz_grid-0009-123456-Cute_cat.png?ex=65f8bbe3&is=65e646e3&hm=4ce801e0326dca6d833aae20165c901b2f6f40233e0d3a692cc9e5563b0a2152&
Here @thorny iron , you can see how the amount of steps affect the output. There is virtually no benefits going above 50... heck even at 35 with Euler A and sd1.5 it's enough. With lcm models you can get descent images at even 3 steps.... turbo models are trained for 1 step generation.

[vestal laurel]

torch not compiled

or maybe my python is not up to date?

idk

[vocal burrow]

Ok so you don't have a proper gpu and only have 8gb of ram..... This is gonna be slow and very limiting.
Still I'd need to see the full console log to diagnose what's wrong with your install.

so copy paste it in a text file and drop said file in there

[thorny iron]

I mean..

[vestal laurel]

i'm really sorry man i'm really a total nobody when it comes to coding i don't understand anything

[thorny iron]

... that 200 stage Meow looks a lot better than the 100 stage one.

[vocal burrow]

...... how so ?

[vestal laurel]

if you can't find the reason don't sweat it i'll try another ywa

way*

[thorny iron]

IT just does!

[open jungle]

/my script add this to requirements to fix the issue, but that is just a workaround

protobuf==3.20.3
insightface==0.7.3

[vocal burrow]

replace those command line arguments with --api --use-cpu all --precision full --no-half --skip-torch-cuda-test

(command line arguments can be found in webui-user.bat)

[open jungle]

replace those command line arguments with --api --use-cpu all --precision full --no-half --skip-torch-cuda-test

[thorny iron]

I agree with Ruoyi here.

[open jungle]

tky

[vocal burrow]

.... Anyways

[thorny iron]

Honestly though, is it actually safe for his Laptop to run this?

[vocal burrow]

Maybe it looks better to you but then that's just how your tastes work. Technically speaking the 200 steps cat doesn't look better than the 50 steps one (or even the 30). Sure when comparing to the 20 steps one we can see that the picture is blurrier.

[thorny iron]

Hmph..

I like his fur better..

[vocal burrow]

sure and you have the right to.

but it's technically not better. It's more pleasing to you.

[thorny iron]

Folds arms :<

[vocal burrow]

That's part of SD random nature, you ask for "cute cat" and it's gonna give you tons of cute cats.... You'll prefer one over the others for sure.

(btw all those were generated with the same seed)

[thorny iron]

200 step cat.

[plucky mulch]

I mean, whether it looks better completely depends on the sampler.

Afaik Euler A technically doesn't converge. Whether it converges more after 30 steps or not, I have no clue

[vocal burrow]

laptops don't catch on fire unless you really push them wayyy too far and put them under bad conditions. They have security to shut down when components reach a certain temperature.

[vocal burrow]

Euler A tends to converge slower than DPM stuff.

[thorny iron]

So I was 99% and it sat there 10 mins I closed the cmd and bsoded wew

[vocal burrow]

But even then it depends of the sampler + the prompt + model + the cfg + etc..... Given the same parameters, Euler A tends to produce less detailed / simplier output. Which can be a good thing if you go for anime stuff for instance. It depends of what you want to generate.

[vocal burrow]

I asked you for a console log trying to help you fix that.

[thorny iron]

Oh.

I'm sure it's because I'm playing Warframe and taking up 4GB of my RAM while doing it. ';d

[vocal burrow]

and to see what settings you've used, because even putting the 150 steps stuff aside there's stuff that could explain this situation.

[thorny iron]

But sure, how do I get a log again?

[vocal burrow]

you generate it again, copy paste what's in the console into a text file (the whole thing), and then drop said file in here

[thorny iron]

Err..

The one that's full of loading bars with "100%" at the end?

[vocal burrow]

the black window with white text in it, yes

[thorny iron]

I'll post it next time it happens, I did BSOD after all. ';d

Also, my text2img isn't SFW :C

[vocal burrow]

yeaaaah test it with a sfw stuff please

[thorny iron]

wew :3

[vocal burrow]

no nsfw content allowed in here

[thorny iron]

Aw :c

[ornate elk]

Yes with --theme dark

Or by setting Windows to dark mode

[ornate elk]

You can get the same cat with 30 steps. More steps dont improve quality, you just get a different image.

[vocal burrow]

don't go there again please :p

I already explained it.

[noble belfry]

mew-ty is in the eye of the beholder?

[high oar]

@karmic crown Sorry to bother you, but would you perhaps know a solution?

[noble belfry]

Not sure what you mean by buttons not doing anything.

which buttons?

do you have a cmd window open for example?

[high oar]

Like, i can't generate any images

NONE of these buttons do anything and no output in terminal

[noble belfry]

it looks like it is attempting to, but is currently trying to interrupt/cancel generating. could you show me a screenshot of the little black window with white text?

(in general, the interrupt process isnt very responsive and I find it infuriating too)

[high oar]

You mean the terminal?

[noble belfry]

yes

[noble belfry]

sorry to sound patronizing, dont know peoples skill level and its the weekend so the range of skill varies greatly

[high oar]

Lol ur good. I ran arch for almost a year before getting my NVIDIA card... it didn't like it very much, so now I'm on mint

[noble belfry]

refresh the browser window, see if it reloads quick or slow. thats normally a good way of seeing if its busy or just stubborn

[high oar]

VERY quick

like 0.5s

[noble belfry]

and theres no change to the terminal window if you press generate?

[high oar]

no, no change

[noble belfry]

very weird. your terminal doesnt do the same thing as the windows one does if you click on it? (freeze and put a cursor where you click and go into select mode)? just take enter on it just to see

if not, close it down, boot it again, server is running but its taking no input and giving no further error or output.

[high oar]

Apologies, don't mean this to be rude if you're not a native English speaker, but I have no idea what you just told me to do.

[noble belfry]

close it, open it again. the whole thing, not the browser window

[high oar]

no difference

wait my VPN wouldn't be interfering, right?

[vocal burrow]

it's all local, so your vpn is unlikely to mess with 127.0.0.1 ip stuff.

[high oar]

yea, that's what i thought, but you never know

[vocal burrow]

I've seen viking handling a couple of "unresponsive" UI stuff.
Generally it seems to be due to some mac/linux setup and/or extensions messing up with dependencies.

[fierce hollow]

What does the CFG Scale affect (In simple words)?

[vocal burrow]

So if you have stuff like mov2mov, deforum, etc... you can check their fixes or wait for them to come back and answer you. In the meantime try turning off such extensions

[vocal burrow]

how much SD has to follow your prompt.

if you have to go above 10 you're probably doing something wrong

[fierce hollow]

Where can I read about SD? I have no idea what it is...

[vocal burrow]

Stable Diffusion

the AI

[high oar]

the extension buttons are unresponsive too

[fierce hollow]

That is, the higher the number, the better the quality will be?
@vocal burrow ?

[vocal burrow]

go into the extensions folder and move the faulty extensions out of it, stable-diffusion-webui/extensions-BACKUP for instance. Then relaunch the UI

[edgy blaze]

Hello! Good morning! Hope you had a good rest. I'm in need of assistance. As I am completely lost now. Last night when I had updated my drives to 24.2.1. Even by doing that. Torch is still not able to use my GPU. Is it because I had completely skipped this step that you said I was able to skip? Or is it something else that I'm not aware of. @ornate elk

[vocal burrow]

No, the higher the number, the more the AI is gonna try to cram every single details of your prompt in the picture.... Even if it makes no sense and it's understanding it totally wrong.

[high oar]

I dont have any in there

[fierce hollow]

I understand you, I guess) And what is the most optimal value? 7?

[ornate elk]

Hey you may need to install the HIP SDK again now, then delete the venv folder and move the DLL files over again

[vocal burrow]

it depends of the prompt and the model. Most of the time the ideal range is indicated in the model's readme/description.

[noble belfry]

im baffled by this. taking no input, giving no output, but working as a server and displaying and loading everything fine

theres no way an adblock would be doing something is there? i mean its local, theres no way it should

[vocal burrow]

lcm and turbo stuff might require a cfg of 1, some models prefer [3;5] some do prefer [5;7]

[edgy blaze]

How do I reinstall the HIP SDK again?

[fierce hollow]

Well, thanks!

[vocal burrow]

it might

[ornate elk]

Just run the installer again
And overwrite

[high oar]

No adblock in that browser

[warm torrent]

Can someone help? Changing models on automatic1111 suddenly freezes the pc, for both .ckpt and .safetensors.

[ornate elk]

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

[noble belfry]

the entire pc?

[vocal burrow]

you don't have anything in stable-diffusion-webui\extensions ?

[warm torrent]

Yep

[high oar]

WAIT, FREEZE. I got rid of the builtin ext and I got an error!

[warm torrent]

Didn't have this problem a day ago

[vocal burrow]

yeah don't get rid of the built-in ones

[high oar]

oh

[vocal burrow]

I have to go for now sorry.

[noble belfry]

try removing --xformers from the webui-user.bat

[high oar]

Well, i'm using the sh, but sure

i got an error this time

[noble belfry]

whichever you're using

[noble belfry]

yeah this is what i was expecting. asyncio errors. dont know why the wouldnt show last time

[warm torrent]

using a 3060 12GB and webui-user.bat args are
@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --xformers
call webui.bat

sorry for late reply, pc froze again while trying to change model again

[ornate elk]

Okay how much ram and free space do you have on C drive?

[warm torrent]

16GB RAM, 100 GB free space on C

[high oar]

What should I do in this case?

[ornate elk]

Are the models 6gb sdxl models?

[noble belfry]

have you edited the sh file?

[warm torrent]

freezes on both 6gb sdxl and sd1.5 2gb models

[ornate elk]

@warm torrent You may need to make your pagefile larger

[high oar]

Other than commenting out the COMMANDARGs and whatever @karmic crown told me to do, no

[warm torrent]

lemme try

paging file size rn is 16384 MB

what should it be after changing

[noble belfry]

what are your command args currently?

[high oar]

-f

but not in the file, just in the terminal

[edgy blaze]

Hey. I think it worked. @ornate elk

[ornate elk]

Set min to 16000 and Max to 24000

And only for C drive

Disable it for any other drive

[warm torrent]

can i set it to another SSD?

it has higher storage

[warm torrent]

or should it be only in the drive that has automatic1111

[edgy blaze]

There seems to be a new problem preventing me from downloading. Do you know how I'm able to resolve this? @ornate elk

[noble belfry]

remove this trailing slash?

[ornate elk]

Yes thats possible

But it should be on a fast ssd

[warm torrent]

Still freezes

Even after changing paging file size

[edgy blaze]

How do I do that? I'm not able to do so in the CMD. Do I have to go and find the file?

[warm torrent]

Freezes during loading weights

Should i try loading it with --lowram?

[noble belfry]

go see if the file is in rocblas\library

[obsidian bear]

Hey guys, I got a question about training Lora with stable cascade. The output I get during training shows a collage of images: [the original image, noised image, prediction, sample, sampled using ema]. Now, the first three look normal while the last two (the sampled) look totally different to what I'm training and seem to just be wild samples. Can someone help me by saying what's going on? Thank you very much! 🙂

[ornate elk]

You need to disable your igpu in the device manager

[obsidian bear]

a bit more context, I'm training a shirt with colours and these show as the last two:

[edgy blaze]

So I have to disable these? When it's finished installing am I able to enable them or do I have to keep them disabled?

[covert shadow]

Hi I'm having this error: torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 20.00 MiB.

GPU 0 has a total capacty of 4.00 GiB of which 0 bytes is free. Of the allocated memory 3.43 GiB is allocated by PyTorch, and 57.55 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

I don't know how to fix

[ornate elk]

No only one

The one with TM

[edgy blaze]

Noted. Thank you, I'll see if that will do the trick.

[ornate elk]

That should do it

We had the same bug and solution yesterday with someone

[covert shadow]

I should have 16 gb ram how do I allocate more to SD

[ornate elk]

Thats an vram issue

WWhat's in Your webui-user.bat?

[vocal burrow]

SD uses vram, not ram

aka your gpu's memory, not your system's one

what's your gpu ?

[edgy blaze]

Victory is mine. @ornate elk

Anyway. Now that I've successfully installed it. Am I able to enable this again? Or do I always have to keep it off.

[vocal burrow]

display memory 3977mb

[vocal burrow]

use --xformers --medvram --no-half-vae command line arguments in your webui-user.bat

[covert shadow]

okay

[warm torrent]

@ornate elk can moving the webui folder to a ssd with a faster write speed to my other ssd help with the crashes from changing models?

[covert shadow]

thank you, this fixed the problem

[paper crag]

so i have the same problem as other people

any suggestions?

Installed Zluda two days ago, worked very well, since yesterday, it gives me this

[ornate elk]

I dont think its SSD related

[ornate elk]

Recheck the install steps

Maybe move the two dll files to torch/lib again

[edgy blaze]

This is meant to take some time right? @ornate elk

[ornate elk]

Yes

Let it load 🙂

15-40 minutes

Just wait until an image is done

[edgy blaze]

Does it matter whether iGPU is disabled or not?

Oh nevermind it loaded something.

[ornate elk]

Nope, doesn't matter

[edgy blaze]

Yeah it's working well now. Thank you.

[ornate elk]

Nice no problem 🙂

[warm torrent]

Fixed my problems by switching to Forge LMAO

I think the automatic1111 commit i used had really bad memory leaks

Thanks for the help

[edgy blaze]

Am I able to do things such as adding Models / Loras somewhere?

[upbeat crane]

type "how add models into auto1111" ... "how to add loras into webUI forge", whatever program you are using. type this into google/youtube etc

there are boatloads of instructions/how-tos for basic usage, i'd be happy to suggest a few creators if you like

[edgy blaze]

Thank you. I understand to some extent now.

[ornate elk]

Models go into the models/Stable-diffusion folder
And loras in models/Loras

[upbeat crane]

check out sebastian kamph on YT maybe. one thing that helps is that once you find a creator or two that you like, if you go to their page, then look at their listing of playlists, you will usually see a date sorted listing of all of their tutorials and such, which is really helpful in order to start building knowledge and taking thigs one step at a time. good luck and have a blast.

[paper crag]

tried it, no difference

[upbeat crane]

they will have a TON of playlists covering a myriad of topics, but you would just want to find a playlist labeled by the name of the program you are using, or maybe sometimes just labeled stable diffusion and you can dig through the titles to find what applies to you

sebastian is good, Olivio Sarikas, i've always enjoyed learning from his videos

[edgy blaze]

Another question. How do I use VAEs? I don't see any drop down menus that allow me to select any VAEs I've installed in the main tab.

Unless I'm blind?

[upbeat crane]

think you have to load that through quick settings in settings by searching VAE

click settings tab, then type quick into the search field

in your quicksettings list, search SD_VAE, then Apply Settings and RELOAD UI

[edgy blaze]

Ah. I understand now.

Thank you all very much. The help is greatly appreciated.

[upbeat crane]

np yw

[night orchid]

how to let Automatic1111 and ComfyUI share models and so on ?

[upbeat crane]

https://www.reddit.com/r/comfyui/comments/14q95sh/sharing_checkpoint_lora_controlnet_upscaler_and/

basically just editing the extra_model_paths.yaml then pulling "example" out of the corresponding lines

then pop in all your pathways

[night orchid]

thanks man

[upbeat crane]

how is Comfy treatin' ya so far, are you using and pre-configured workflows?

i've grabbed Sytan's SDXL workflow and SEARGE'S SDXL evolved so far got some nodes updated so far

but havent dug too dep as of yet but soon

[night orchid]

I am a beginner when it comes to ComfyUI

[edgy blaze]

Is this some form of glitch? It's saying I have no memory available to do a certain resolution, when I do indeed have enough memory.

[night orchid]

no I think it is your Vram its complaining about

SD uses mostly Vram

you could use CPU but its not recommended, because it is very slow

[edgy blaze]

How do I clean that up then? I have almost no tabs open and no apps open rn

[upbeat crane]

it's working fine

it will ebb and flow, go up and down and all over the place depending on your values and what your doin

[night orchid]

@edgy blaze no no need to clean up, you just need enough of it

what is you graphic card?

[upbeat crane]

i also have 16gb of vram on my GPU and i peg it out and get out of memories at times

[upbeat crane]

i mean, if it's stuck there its an issue

[edgy blaze]

Yeah. Definetly.

[upbeat crane]

what resolution are you trying to push btw?

[night orchid]

you have 8GB wich is good in most cases

[edgy blaze]

512x768. I max out at 512x512 strangely enough.

[upbeat crane]

yeah i wouldnt worry about monitoring your VRAM just based on numbers in your head