🤝｜tech-support

[ornate ruin]

Can you figure out the used hires fix from a PNG data? Everything except for that seems to be transmitted via the PNG Info tab

Without that it does seem to be impossible to reproduce an exact image?

[tranquil cove]

Hi. I opened webui-user and set PYTHON="C:\Users\David King\AppData..." but when I try to run it it says "No Python at '"C:\Python311\python.exe '. How do I get it to recognize the correct python path?

[ornate elk]

Hey, try delete the venv folder and rerun the webui-user.bat

[tardy sparrow]

Does anyone knows how to use the basic plan of modelslab.com to generate image faster on A1111 ?

[ornate elk]

I dont understand the question

Settings, User Interface, Quicksettings, there add facerestore
Then apply and reload

[paper crag]

hi guys, new guy here trying to make stablediffusion work. I'm at the point of running the webui-user, and its not working.
Apperently i found out the regular stablediffusion files you download are made for Nvidia. I have AMD. Do i have to make a ton of changes, or do i just replace the webui file with a version that works with AMD?

[ornate elk]

Hey, for AMD you need a different webui version.
Checkout my install guide for AMD. Its in the Pinned Messages of this channel

And because you downloaded the nvidia version before, you need to do the Important step mentioned there too

[paper crag]

so before i attempt any other changes, i should go and input 'pip cache purge' command first right?

[ornate elk]

Yes

[paper crag]

okay, done. The 34 files were removed. Now should i just jump straight to step 5? Or start over the AI file?

[ornate elk]

Did you git cloned the new repo?
If so then yes step 5

[tranquil cove]

That worked thanks! Working through the rest of the install process now.

[ornate elk]

Nice, checkout my guide in the Pinned Messages for the best performance and compatibility settings

[tranquil cove]

Thanks! Not there yet. I closed the terminal before copy/pasting the URL and when I try to relaunch it, it says "Torch is not able to use GPU"

[paper crag]

the git i cloned was this one
https://github.com/AUTOMATIC1111/stable-diffusion-webui

[ornate elk]

What's your GPU?

[ornate elk]

Pls checkout my guide

Like I said you need a different version for amd

You can delete the stuff you downloaded before

[tranquil cove]

AMD Radeon RX 6950 XT

[paper crag]

MSI Gaming Radeon RX 6800 16GB GDDR6

[paper crag]

okay

[wooden iron]

i am also using amd gpu , comfyui

if you need help ask me

[ornate elk]

Then you need to also checkout my install guide for AMD, because you downloaded the wrong version

Also make sure to do the important step mentioned there

[wooden iron]

are they instaling comfyui or automatic 1111

[ornate elk]

Automatic1111

[wooden iron]

use comfyui

[ornate elk]

Nah wouldn't recommend for starters

[wooden iron]

hard to understand but better control to image

[ornate elk]

Depends on what you want to do

But yea it has more settings to tweak for sure

[wooden iron]

i heard comfyui faster then automatic 11111

mir8

[ornate elk]

Yea for nvidia that's true.
Dont know about directml

[wooden iron]

bro i am having problem

when i use reactore face swap

you can see face is slightly pixalated

how to solve this

[paper crag]

before i go and install the wrong version, if my end goal is to use stablediffusion for anime fanarts, whats better, comfyui or automatic 11111?

[ornate elk]

You can do that with both, but Automatic1111 is better for beginners, has more extension support and more people which you can ask for questions

[paper crag]

automatic it is then

lets see if this works

[ornate elk]

You also can install both, so its not a lifetime choice xD

[paper crag]

righhhhhhht

[ornate elk]

Upscaling could fix it maybe

[ornate elk]

Should work ^^
Used it on a 6700xt, 6800xt, Vega7, and 7900XTX

[wooden iron]

i am using comfyui bcz its my choice

and i wont regret a bit

boo automatic 11111

https://tenor.com/view/determined-anime-gif-7558450

[ornate elk]

What was your GPU again?

[wooden iron]

rx6750xt

12gb

[ornate elk]

Ahh nice card

[ornate elk]

Is your controlnet updated?

That folder path looks strange

[vocal burrow]

You ve put a .bin in there instead of .pth or .safetensors

[grizzled mural]

https://github.com/lllyasviel/ControlNet

https://github.com/dreamscapeai/stable-diffusion-webui

[ornate elk]

Thats okay for IP-Adapter face Id, It will work

[grizzled mural]

i changed the name from bin to pth, but then i also need the preprocessing

[vocal burrow]

Ok then I ll go back to sleep :p

[paper crag]

welp, i just screwed over by luck. Power fell when the downloading was at 89%....
Hope this won't cause any issues...

[ornate elk]

Depends on what it was downloading

If it was the 1.5 ema safetensor model your fine

[ornate elk]

Yea thats okay

Its just the base model

[paper crag]

yeah, sorry if im to paranoid

[ornate elk]

No no problem at all

[paper crag]

i've been trying to make it work the past couple of days and when i finally solve a issue i encouter, i hit a new one

[ornate elk]

Yea that can be frustrating, so much outdated and bad tutorials out there

Even the "official" sd amd guide isnt a good help xD

[paper crag]

yeah, out of the dozens make it work tutorials i watched, no one warned me 'Make sure you check what GPU you have before you struggle like a idiot'

[ornate elk]

This webui is 6 months outdated, I wouldn't recommend using that over the normal automatic1111 webui.

[ornate elk]

Yea thats so annoying xD

[grizzled mural]

idk why the guy on the tutorial put that custom webui...

https://github.com/AUTOMATIC1111/stable-diffusion-webui this is the official one , right?

[ornate elk]

Yea it hasn't any features the normal automatic1111 doesn't have. But has less updates xD

@grizzled mural Checkout my install guide in the Pinned Messages of this channel. There is everything you need to know

And yes that link is the normal one

[grizzled mural]

im running on sagemaker studio lab, so limited storage

using a tmp for checkpoint models

[ornate elk]

Ah okay

[paper crag]

okayyyyyy

so it seems we are in business

lets see what this thing can do

[ornate elk]

Cat images and anime waifus

For Anime stuff you need to download an Anime model from civitai.com

[paper crag]

I am aware, i need to set up a loooooooot of stuff before i can generate some decent pics

[ornate elk]

Yea not to much but its a learning process ^^

Let me know if you have any specific questions, also this Directml Version is not as Vram optimized as the nvidia version.
So keep in mind when you get an Out of memory error that you always have to restart the webui console to clear the stuck vram.

[patent meadow]

i followed a couple of suggestions and got here

should i start from scratch

[ornate elk]

You should checkout my install guide for AMD, its in the pinned messages of this channel.
In fact just remove the --onnx --reinstall-torch

Then add --medvram --opt-sub-quad-attention --opt-split-attention-v1 --upcast-sampling --no-half-vae

Then save.
Then delete the venv folder.
And launch the webui-user.bat

[patent meadow]

i have 16gb of vram should i do it anyway?

[ornate elk]

Yes

[patent meadow]

ok

[ornate elk]

Directml isnt vram optimized so its needed

[patent meadow]

i should just re run webui-user.bat?

[ornate elk]

That was an internet connection error

Relaunch the webui-user.bat

[patent meadow]

thanks foe clarifying

[ornate elk]

Make sure you have a good connection

[patent meadow]

slow internet in iraq

[ornate elk]

Ah damn :/

[ornate elk]

If you have a model already you can put it into the models/Stable-diffusion folder.
That will prevent that the base 4.5gb model will get downloaded again

[ornate elk]

You have to use the venv command first

When in the folder use:
py -3.10 -m venv venv

venv\Scripts\activate

And then the run.py

[wind sun]

thats never a good thing

[ornate elk]

Type cd facefusion

Not face fusion

[wind sun]

opps

true

[ornate elk]

You should type it in. Not click on it

[pearl musk]

Hey guys, is openpose working? I wanted to download it but the repo is read-only

[ornate elk]

Hey, there is no openpose repo to download. As its only a model you need for the Controlnet extension to get it to work

[pearl musk]

Oh, thanks a lot for the quick reply

[ornate elk]

Np, After installing the controlnet extension, you can get all needed models from here.
https://civitai.com/models/38784?modelVersionId=44811

[wind sun]

thank you

you came in clutch

like aways

❤️

[ornate elk]

No problem 😁

[ornate elk]

For launching it again you only need to do the
Venv activate command and the runpy

Not the python 3.10 -m

[wind sun]

i was just gonna ask

[ornate elk]

I'm off now gn 😴

[patent meadow]

good night 🌃

[sonic ferry]

dumb question, SD can't work with a GPU installed in one of those crypto mining adapters that are connected to USB, can it?

[jolly pawn]

error happens when I run ip adapter preprocessor

[crude pecan]

Hey everyone, I recently launched Auto 1111 SDK (https://github.com/saketh12/Auto1111SDK), an open source, lightweight, modular Python library for the Automatic 1111 Stable Diffusion Web UI. I would appreciate if you all could check it out and give it a star!

[ornate elk]

Hey, awesome project!
Is the token lenght limit of 77 also disabled like in auto1111?
If so you should add that to the feature list 🙂

[ornate elk]

You installed controlnet the wrong way by downloading it as zip file

Normaly you should install it via the Extensions tab.

It seems like the clip file got corrupted

[worn frigate]

hi guys please help
I have downloaded control net models but they seem to have 1.5 in their name, and after they did not work i found i have this version of stable diffusion install which is newer version i guess?
I am trying to use control net and its not working what can i do pls help , also is my place to paste control net models right?

[ornate elk]

Hey, Webui 1.7.0 is the latest version

You need to update with git pull

[worn frigate]

will it work afterwards or i will have to install controlnet's different models? are there like sd16 sd17 or we use 15 for all

thanks for responsding i appreciate it

what do i need to add for git pull?

[ornate elk]

After updating I would recommend to download the controlnet models + their config files from here:
https://civitai.com/models/38784/controlnet-11-models

[ornate elk]

What's your GPU?

[worn frigate]

nvdia rtx 3070 laptop

[worn frigate]

wil the models i have installed not work?
they are 1.5 gb in size and i have like 9 of them 😢
can i go back to sd15 where these work.

I really dont know please guide me

[ornate elk]

okay then replace the whole text in the webui-user.bat with this:
`@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--xformers --medvram-sdxl --no-half-vae --api --update-check
git pull
call webui.bat
`

[worn frigate]

done

[ornate elk]

thats not done lol

[worn frigate]

will i have to remove max split size mb 128

[ornate elk]

yes

and add --xformers --medvram-sdxl --no-half-vae

[worn frigate]

oh..

can i add it back after words?

the 128 mb line

[ornate elk]

i dont understand that question sry

[worn frigate]

sorry if i am talking non sense but there is model call juggernaut xl and it does not work without this 128 mb line because it says the pytorch cuda needs more memory

[worn frigate]

this line

[ornate elk]

the problem is that you didnt used --xformers --medvram-sdxl --no-half-vae

there is no need for cuda alloc mb128 stuff

[worn frigate]

ok i am pasting it

[ornate elk]

it will be faster with my stuff

so remove the whole set pytorch alloc cuda mb128 line

[worn frigate]

ok thanks luffy

[ornate elk]

no problem, show me the file again after you edited it

[worn frigate]

i am afraid i will have to go with my little sister

and i will be back after a while

then i will do it

so sorry and thanks you are the best

[ornate elk]

no problem 🙂

[worn frigate]

I am back. Sorry this might sound different. But will doing it cost me data more than 2 Gb? There is a problem that here in my country internet is kinda expensive. And most people use cellular data plans. And I get only 2GB per day. I do not have a wifi. But they have a offer which provides Unlimited data from midnight 12 to 6 am. Because most people are sleeping and it does not cost much to sim card company... So if that is the case i will have to run this web ui bat at that time. This is the reason i have git pull removed from it in case its like 8Gb file. And that is why I was asking if i could make use of those 1.5 canny depth etc models.... Because another thing is I am kinda younger and my dad does not allow me to use PC specially at night.. I have to sleep early for school.. sometimes I be sneaky and use it when he's sleeping.. but its not possible daily.. thats why i was hesitating to make any changes to it in case It needs to download big file to work and i wont be able to use SD until full files download. So i thought maybe there is a way to go back to 1.5 so these models work.... I can use my phone at night.. and then i send models to PC during day... I am sorry for such long problem thing... Will it consume too much data?

[ornate elk]

the git pull wont be more than a few hundred MB not GB.
and yes you can use the controlnet models your already downloaded.

[worn frigate]

Thank you, i will try to run it as soon i can. I will update you with it. Thanks : )

😄

( i am on my phone rn)

[ornate elk]

np, if the controlnet models dont get recognised you can try and move them into models/controlnet folder

[worn frigate]

Thank you

You help almost evryone here. That is so rare personality nowadays.

You are so cool😎

[ornate elk]

Thanks! i like to help people to get a good experience when using Stable Diffusion 🙂

[little bough]

Hi, is anyone using the new forged a1111 by lllyasviel? I've installed it and liking it but having an issue with LCM, it seem that because I have Animatediff installed which has the LCM sampler aswell, that I now have 2 in my list but the real issue is I can't seem to use the LCM LORA as I get a Tensor error as if the 2 LCM's are clashing, the native LCM that has come with his model and the LCM that comes with animatediff, does anyone know what I'm referring to and a way to say get rid of the Animatediff LCM? Thanks. I've updated everything

[tender gazelle]

Does anyone from Stability know why the stable-zero-123 channel was archived? (It doesn't get a lot of activity, but that's because Stability built on a pipeline that doesn't support Windows. It's still the only place most people can go for the slim chance of getting it working.)

[fair oxide]

Could you tell me how to use img2img? For example, how do I use a selfie and generate an image of a warrior with my selfie?

[zinc pulsar]

I have this strange problem that my Automatic1111 cant reproduce old images i made. Pics are similar but always worse, as the resulting image seams to have way worse anatomic features. For example the stomach area is distorted, or way to muscular. In addition, every new picture i create hast the same issues. They all look descent but not as good as prior and especially with some huge errors.

[elder turtle]

Got examples for new and old ones? Also what timeframe were talking? I think something about the seeds changed ever since the start of A1111

[zinc pulsar]

Pic on the left is from November, right one is new. As you can tell the navel is just a black hole in the clothing in the right one. Prompts and everything else is the same way as in the left one.

[tender gazelle]

Since the images you posted don't have parameters saved in them, no one can help you. Sorry.

[zinc pulsar]

thats a discort problem give me a sec.

[tender gazelle]

Maybe you can post the parameters / screenshot them from A1111?

[tender gazelle]

Yeah, I'm not downloading a random zip file, sorry. (Not safe.) Gotta go to work. Good luck though.

[zinc pulsar]

i get that.

[elder turtle]

Also discord doesnt strip the metadata I think

[zinc pulsar]

It has to, these are my original files.

[ornate elk]

Nope discord doesn't strip Meta data

[zinc pulsar]

I get it, you guys dont trust me lol

[fair oxide]

It's safe. I downloaded and scanned it with antivirus.

[zinc pulsar]

there i took screenshot

[fair oxide]

Does stability gprc api has rate limit?I can only see rate limit on rest API in documentation.

[zinc pulsar]

and yes i already saw a diffrence compairing it that way. Still does not explain why any of my new pics have distorted bodys sometimes, but unfortunately i dont have examples since i delete them as soon as i get them. I am currently trying to get a example

[worn frigate]

so now i run it right

i dont know where is venv directory and git.

do i leave them empty?🙂

also I do not have xformers i think because it usually says no xformers starting without it.

I have a GIt files named folder though, do i need to add its location there?

and a folder named venv too

what do i do

@ornate elk

[zinc pulsar]

Appreciated the time you guys took. I had this problem for months and now i have no ide what i did, but i can reproduce my old images again.... Would love to know how but well at least its fixed for now. Thanks for the help no less.

Is there a way to save a current state of SD? all settings and stuff?

[worn frigate]

So i moved the controlnet models to models/controlnet and its working now.
also added git pull to the webui it updates to 17

i did not know about other stuff so i left unchanged 😕

i am trying to install comgy ui now

`@echo off

set PYTHON=C:\Users\Dipper\AppData\Local\Programs\Python\Python310\python.exe
set PYTORCH_CUDA_ALLOC_CONF=garbage_collection_threshold:0.6,max_split_size_mb:128
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --api
git pull

call webui.bat`

its what it looks like. its probably wrong like you said.

[worn frigate]

i am running this now

its installing xformers.

[pearl musk]

Hey guys, can someone explain Hypertile to me? I am currently stuck on a GTX 960M GPU with a Core i7 (I know, but this is my personal laptop I'm very fond of) and gens with Adetailer at 578x768 take roughly 4 minutes lmao.

[tender gazelle]

1. Maybe the new adetailer version switched to using a different sampler by default? I don't use it, so I don't know.
2. Are you using xformers? Because xformers will give you slight differences for each generation, even w/ fixed seed.
3. Maybe A1111 switched to GPU sampling or something? It's been available in comfy for a while. That could change sampling output.

[opaque viper]

Questions please. I have a 1660S (6 GB VRAM) so I'm trying out Forge to run SDXL models. The speed is much better though still slow (about 6-6.5s/it). I heard using fp16 can maybe speed up the generations for some GPUs, so I want to try that. But I'm not sure which of these arguments I should use:

Just all in?

Tried using all in fp16 and it's slower now

[vocal burrow]

When talking about fp16 90% of the time people mean "use fp16 model instead of fp32". SDXL base model is already fp16. So there's no point doing any of that.

[opaque viper]

Ah okay, thanks for the clarification

One other thing. Is this how I do it if I want Forge to use A1111 directory for my stuff?

set PYTHON=
set GIT=
set A1111_HOME=C:\stable-diffusion-webui
set VENV_DIR=%A1111_HOME%/venv
set COMMANDLINE_ARGS=--ckpt-dir %A1111_HOME%/models/Stable-diffusion^ --hypernetwork-dir %A1111_HOME%/models/hypernetworks^ --embeddings-dir %A1111_HOME%/models/embeddings^ --lora-dir %A1111_HOME%/models/Lora

[vocal burrow]

@zinc pulsar adding this to the list of potential culprit https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Seed-breaking-changes

[vocal burrow]

you're missing the VAE and gfpgan directory (and probably a bunch of other for your extensions such as controlnet)

I don't know about forge's specificities

[opaque viper]

What I did was what was written in the webui-user bat file, but it doesn't seem to be working. I'll try out other stuff

[ornate elk]

Hey, how does your file look now?

[worn frigate]

It looks like this now.

Looks like i will have to download and put styles.csv there next time i get pc.

With python =its location

[ornate elk]

Okay good.
Styles.csv isnt needed. It will get created when you save a style

[worn frigate]

But i would not launch without it?

And without launch how would I save a style.

[ornate elk]

It should launch

Do you get any error?

[worn frigate]

No but it did not progress after it. I didnt retry as i had to close pc.

It will work ig.

[ornate elk]

Yes it should work

[worn frigate]

Thanks

[tardy sparrow]

hi, anyone know if its possible to use https://modelslab.com/ api/gpu with A1111 to generate my images faster ?

and if yes, how ?

[ornate elk]

hey, whats your gpu?

and yes you can use the api o any other cloud gpu rent service to generate images faster

[tardy sparrow]

i have rtx gforce 2080

[tardy sparrow]

i bought basic plan thinking it was easy to configure...i cant for the life of me understand how it works

[tardy sparrow]

they say they have a dedicated GPU and i have 3600 api request availables now

[ornate elk]

ah okay, for the best performance with this gpu you need to add --xformers --medvram-sdxl --no-half-vae
to the wbeui-user.bat

[tardy sparrow]

i've tried --xformers --no-half-vae before

the image generation result was slighty different, and i didnt liked it

so i deleted this

[ornate elk]

these arguments doesnt change the output quality

[tardy sparrow]

oh

really ?

thats weird

[ornate elk]

so no need to delete the best performance boost you can get xD

[tardy sparrow]

you are absolutely sure about that ?

because i have an example with / without

for the same seed

same configs

[ornate elk]

yes im sure about that

[tardy sparrow]

mmm

okay well

i'll give it another try then

[tardy sparrow]

should i use that with A1111 ? is that even possible ? how do i say to the software to use a dedicated gpu instead of mine ? would you know any good tutorial ?

[ornate elk]

idk how to integrated an api, you should have bought an virtual vm with a cloud gpu, from services like: runpod.io, paperspace, vast.ai for example

there you would have to install a seperate auto1111 inside the VM

[tardy sparrow]

fak me

well

thank you anyway

[ornate elk]

np, but your gpu is powerfull enough that you dont need any of this to use SD with a decent speed

but --xformers --medvram-sdxl --no-half-vae is needed for that

[tardy sparrow]

i just tried

i don't think the difference of speed is really noticable

but my output is different

for sure

without

with command argument

very slighlty difference

i agree

[ornate elk]

i tested it too right now

left without and right with xformers

[tardy sparrow]

thats weird...

yeah well

u have some differences too

very slighty too

less than me tho

i know that the same seed can give differents results ofc

maybe it's just that

[ornate elk]

these images dont have the same meta data

[tardy sparrow]

everything good with my versions ?

[ornate elk]

so this is not a valid test

[tardy sparrow]

mines ?

uh

[ornate elk]

yes

[tardy sparrow]

okay my bad then

[ornate elk]

they have different settings

[tardy sparrow]

i must have fckd up

smth

well thank u tho

i'll keep the command arg line

btw

when i add a new lora

[ornate elk]

i check the meta data and there was a change in the prompt:

[tardy sparrow]

wait what

wtf

[ornate elk]

seems like a different lora got used, or you renamed a lora mabye

[tardy sparrow]

that should not have happened

i'm supposed to use race twilek lora then race twilek v2 lora

how the fuck did that changed

anyway,

bien vu

well done

wait

why

my metadata is like this

but when i drag the image into A1111

it becomes this :

why is Race - TwilekV2 changed into twilek

[ornate elk]

whats the filename of them?

[tardy sparrow]

00107-3395022432

what the shit

[ornate elk]

thats strange

maybe a corrupt file?

[tardy sparrow]

mmhmhh maybe

or maybe its because

mh wait

its happening again with another file

i think its because

the same lora is used 2 times

but with 2 differents versions

so when i drag the pic into A1111

something must be fucked up

and he doesnt understand that i'm using the 2 versions at the same time

[zinc pulsar]

Sounds similar to my case. I have been trying for the past hour to recreate what i did to fix my images. I somehow got it to creat the same pic as the november one, but now i can not recreate the pic i got yesterday. Then i switched around my modules, and something must have changed because i could recreate my pic from yesterday, but at the same time the pic from November was gone again... Going crazy hear.

[tardy sparrow]

Do you use differents versions of the same LOA too ? Did you add extensions since then ?

[ornate elk]

the lora file for sure if the name are just numbers

[tardy sparrow]

it's normal to not recreate the exact same image, but differences should be minimal

[tardy sparrow]

Can you try drag my "00107-3395022432" image in your A1111 and tell me if you have

in your prompt ?

just to make sur the issue isnt only on my end

[zinc pulsar]

You would think that, but why can i recreate the same image every day for 2 monthes, but all of a sudden i can only recreate the version of that image I got 2 monthes ago but not the version i got every day since then?

[tardy sparrow]

Idk ,tbh i found myself in the same situation

sometimes i recreate the same perfect image

sometimes not

check if your pic from 2 month metadata matches the prompt / negative / configs u have now

maybe u have the same problem problem as me, meaning the metadata does not copy correctly into A1111

[ornate elk]

thats in png info

[tardy sparrow]

Ok so the png info is correct

[ornate elk]

when i send it to txt2img

[tardy sparrow]

WAT

what the hell

so for you, both png info and text2img are corrects

but for me when i send it to text2img it's being changed

thats creepy

[zinc pulsar]

left to right, 1. pic i got yesterday. 2. pic after i was able to recreate the pic from novermber i asked you guys prior, 3. pic after i tried to recreat the pic rom yesterday, 4. the pic i got now. (i hope the meta data is not missing again)

[tardy sparrow]

@zinc pulsar metadata missing, indeed

btw

when i send it to txt2img, without fixing the lora prompt

when i send it to txt2img, fixing the lora prompt for twilek - V2

so i guess you have to be careful when using 2 versions of the same lora

[zinc pulsar]

this is so stupid, i see the meta data in the comment section of the file and in png info. Why are they not visible in discort....

[tardy sparrow]

i have no idea

but since u can see the metadata urself, pay attention if the metadata is the same between the 4 images, and then, when you drag them into text2img, if the metadata is still stays the same

[ornate elk]

yea i didnt had these case before

because nobody uses two of the same loras

[tardy sparrow]

i'm nobody

😂

well as u can see

i have noticable better results

when i mix them up

oh by the way @ornate elk last question i hope

when i add a lora to A111 while A1111 is running, should i close it completly then re-open it ? or just reload UI ?

[vocal burrow]

click the "refresh" button

[ornate elk]

at the lora selection is a reload button

same goes for the models

no need for a restart

[tardy sparrow]

okay thank u

just wanted to be sure

[worn frigate]

Um
What does -medvram do?

Is there also like -lowvram and -highvram.

What does it change and what doing -medvram on 8gb vram do?

🤔

[tardy sparrow]

@worn frigate Makes the Stable Diffusion model consume less VRAM by splitting it into three parts - cond (for transforming text into numerical representation), first_stage (for converting a picture into latent space and back), and unet (for actual denoising of latent space) and making it so that only one is in VRAM at all times, sending others to CPU RAM. Lowers performance, but only by a bit - except if live previews are enabled.

[ornate elk]

you need --medvram-sdxl

--medvram is only for 6gb vram or less

[worn frigate]

Thanks, i have medvram and sdxl which you wrote

[timber oak]

hey guys

[worn frigate]

: )

[timber oak]

git pull
@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--xformers --no-half-vae --precision full --no-half
set PYTORCH_CUDA_ALLOC_CONF= max_split_size_mb:512

call webui.bat

i was told there are ways to improve my webui-user.bat launch settings

[ornate elk]

yes, whats your GPU?

[timber oak]

rtx 4070

[worn frigate]

Wow. Thanks so much.

If i have all on one vram.
Like if i do not add this line.
What am i doing?

[timber oak]

changed from a 1060

[ornate elk]

wow, you slowed your self down a lot with that arguments

[timber oak]

haha

[tardy sparrow]

No idea, i just copy pasted what the wiki ( https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Optimizations) said

;D

[timber oak]

so

do you mind telling me what these do?

[worn frigate]

You are just like me XD

[ornate elk]

remove the set Pytroch Cuda alloc 128mb stuff line, completely,
then remove --no-half --precision full
and add: --no-half-vae

[timber oak]

my friend sent those to me some months ago when i had my 1060 bc it kept crashing trying to generate images

[ornate elk]

yes these are okay with a 1060 but not for a 4070

[timber oak]

git pull
@echo off

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

call webui.bat

good?

[ornate elk]

yes

[timber oak]

anyhing else?

[ornate elk]

nope

[timber oak]

id love to know what these changes actually did ^^

[timber oak]

what do these do

[worn frigate]

Offtopic but how much 4090 , 80 or 70 go for now.

I checked and 4090 was like $2500.

[timber oak]

bought my 4070 for 650

eur

[tardy sparrow]

damn it

[worn frigate]

$650 dollar?

How much american dollar is that

[timber oak]

so maybe like 700-750 dollar

[tardy sparrow]

u guys have 4070

while i have 2080

i'm jealous

[worn frigate]

I dont have.

[tardy sparrow]

😦

[timber oak]

2080 also sexy

[tardy sparrow]

true

[worn frigate]

Ye

2080 so cool

[tardy sparrow]

:3

[timber oak]

1060 is my favourite granny tho

[tardy sparrow]

TRUE

1060 was the bomb

[worn frigate]

How much do you believe 4090 is?

Just curious.

About the prices.

[ornate elk]

xformers is a performance boost for nvidia cards, it uses less vram and speeds up generation time.
--no-half-vae is needed for vae files that are used for color correction with models like sdxl vae or anime vae

[timber oak]

i have no idea what a vae does

never used it

but sometimes when i make a prompt too long the colours bleach out

[zinc pulsar]

thats what a vae is there for

[timber oak]

aah

epic

i didnt know that

[zinc pulsar]

in my experiance its enough to get one for real and one for anime. No need for the special one the model you use wants. But it can help

I only use 2 vae and i am happy with that.

[timber oak]

ok thx

is it like a lora?

like

in the prompt

[zinc pulsar]

more like the checkpoint. You set that in advance

[timber oak]

ok thx

[tardy sparrow]

stupid question : what does embeddings means ?

extensions ? lora ? models ? all of that ?

[zinc pulsar]

from what i get, its similar to a lora. Its a word that links to a Text file that tells SD what do do.

For example Badhands, is an embedding I use, thats just a faster way to get all the negative hand prompts without having to write them.

[tardy sparrow]

aaah

[zinc pulsar]

you can also get some Characters as Embeddings instead of a Lora.

[tardy sparrow]

okay

i see

well i use my own styles

i prefer it that way

thanks tho !

[ornate elk]

vae

None means no custom vae used

[faint mica]

trying to do a fresh install and getting the error:
stderr: ERROR: Could not install packages due to an OSError: [WinError 5] Access is denied: 'C:\Users\User\stable-diffusion-webui-directml\venv\Lib\site-packages\onnxruntime\capi\onnxruntime_providers_shared.dll'
Check the permissions.

[ornate elk]

hey, checkout my install guide for AMD. its in the pinned Messages of this channel,
basicly you have to make new folder first before starting the git clone into it

[worn frigate]

Where can i choose vae?

[ornate elk]

you need the vae dropdown. to get that go into Settings, User Interface, Quicksettings, there add sd_vae
then hit apply and reload

[worn frigate]

Thanks.
Are theese vae downloaded? The ones in picture

[ornate elk]

yes, you need to download them and then put them into models/vae

[worn frigate]

Or they built in stable diffusion as soon as i turn setting on

Ty

Also, Is a SDXL model good thing or too much pressure on pc?

I see some models are XL,
Like juggernaut XL.
So that means they are trained on higher resolution images and needs more power?

Is that too much for a rtx 3070

I saw that there is XL for controlnet models too.
Its 2.1 or similar.

I do not know terms much srym

[ornate elk]

yes sdxl models are trained on 1024x1024 and need more vram, but your fine with 8gb and the --xformers -medvram-sdxl will speed that up

and yes for sdxl models there are also controlnet models

[worn frigate]

Ty.

Also um,
How fast is comfyui compared to webui?

On average. How much percent fast it is maybe.

110%?
150%?

[ornate elk]

maybe 10%

or 15%

[worn frigate]

Hm

[tardy sparrow]

Guys

Do you know a easy way to change some part of the image i generated into back ground ? with img2img or inpaint for example

Like sometimes i have an extra arm or a part of body that is not supposed to be here, i would like to replace it with background

[covert jolt]

m

[tardy sparrow]

m

yes

[worn frigate]

If you want to remove something from image.

Best bet would be 'Lama painter'

[tardy sparrow]

lama painter ?

[worn frigate]

Ye

[tardy sparrow]

gonna take a look at it

ty

[worn frigate]

Its epic.

[tardy sparrow]

ok i don't find it

what it is XD

[worn frigate]

Its an extension

It allows you to paint with brush like inpainting

And it removes whatever you paint

You have a chair in image and want to get rid of?

[tardy sparrow]

hem

it's nsfw

[worn frigate]

Use your brush and it will be gone

No ,not nsfw

Lama painter is extension.

For inpainting.

[tardy sparrow]

okay

i cant find it tho : (

[worn frigate]

Its called lama painter.

[tardy sparrow]

isnt it lama cleaner ?

do you have the link of the git hub ?

please

[worn frigate]

Oh wait

Yes

So sorry

[tardy sparrow]

hahaha

no problem

[worn frigate]

Its lama cleaner

Lel

It slipped from brain because we paint

So i thought its lama painter.

[tardy sparrow]

https://github.com/Sanster/IOPaint?utm_source=100xgrowth.ai&utm_medium=marketplace&utm_campaign=100xgrowth.ai

is it this one ?

[worn frigate]

Idk

I do not understand files much either.
They all look same

[tardy sparrow]

https://github.com/novitalabs/sd-webui-cleaner?tab=readme-ov-file

i found this one, looks amazing

[worn frigate]

I have no idea which one is correct

I am not on pc otherwise would have sent link.

Someone had sent link to me and i do not remember which one was it.

I just copied pasted.

[tardy sparrow]

okay no problem

thank u tho !

[faint mica]

so that worked but now I'm getting this weird color when trying to inpaint. This just started today after the recent update. The preview while the image in generating looks fine but the result looks like this.

[ornate elk]

For getting inpainting to work with AMD GPUs you need to add --no-half to the webui-user.bat.
Keep in mind that this will result in a slowdown in generating speed and more vram usage

[faint mica]

yup that fixed it. Thank you

[timber oak]

yooo thats such a difference!

[rare cradle]

Hi all, i have a problem using stable diddusion and Loras. I can't see some of them in the u.i. when I check this they appear but unuseable.

i seached all solution in the web but nothing works.

Sry for my english

[analog shore]

I'm sorry to bother again with this things but I've problems with CUDA because I try to install Wave2Sync and prolly I've messedup the CUDA, so I've unistalled it and resinstalled the CUDA toolkit 11.8.
I've also reinstalled A1111 but I still get the error:

  File "F:\AI\stable-diffusion-webui\launch.py", line 48, in <module>
    main()
  File "F:\AI\stable-diffusion-webui\launch.py", line 39, in main
    prepare_environment()
  File "F:\AI\stable-diffusion-webui\modules\launch_utils.py", line 384, 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

So, how can I fix it?

[tardy sparrow]

did you try checking that box ? what version of A1111 do you have ? python version ?

[analog shore]

Please 🥺 🙏

[rare cradle]

yeah, like i said, if i check it they appears but i have an error when i use them

[ornate elk]

Change the model and then they will show up
Loras are for 1.5 or sdxl.
So they only show up when selecting the right model version

[rare cradle]

ok

it works, i didnt know the models affects lora, i cant see them all but the problem seems coming of that, thanks a lot 🙂

[tardy sparrow]

thats interesting, i dint knew that

[ornate elk]

No problem 🙂

[tardy sparrow]

i thought all models were 1.5 or sdxl

[ornate elk]

There are also 2.1 models and loras

But not used that much

[tardy sparrow]

lets say all my prompts and configs are good, but i want more variation in the output

what else than seed -1 can i change ?

[analog shore]

Ok looks like I've fixed it... thanks to @analog shore
To fix CUDA remove and reinstall CUDA Toolkit 11.8
pip install torch==2.0.1+cu118 torchvision torchaudio —extra-index-url https://download.pytorch.org/whl/cu118

delete the Venv folder
launch A1111 let it download all the needs.

😎

[sweet basin]

new sd update today and it doesnt work anymore

E:\stable\sd\webui\stable-diffusion-webui-directml\venv\lib\site-packages\pytorch_lightning\utilities\distributed.py:258: LightningDeprecationWarning: pytorch_lightning.utilities.distributed.rank_zero_only has been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it from pytorch_lightning.utilities instead.
rank_zero_deprecation(
Traceback (most recent call last):
File "E:\stable\sd\webui\stable-diffusion-webui-directml\launch.py", line 48, in <module>
main()
File "E:\stable\sd\webui\stable-diffusion-webui-directml\launch.py", line 39, in main
prepare_environment()
File "E:\stable\sd\webui\stable-diffusion-webui-directml\modules\launch_utils.py", line 641, in prepare_environment
from modules.onnx_impl import initialize_olive
File "E:\stable\sd\webui\stable-diffusion-webui-directml\modules\onnx_impl_init_.py", line 7, in <module>
import optimum.onnxruntime
ModuleNotFoundError: No module named 'optimum'
Press any key to continue . . .

@ornate elk

[formal trout]

Do you know why I might be getting two images (in the folder) after one generation while inpainting? I just notice one is a jpg and the other png

[ornate elk]

there wasnt an update for me

[sweet basin]

it was big

a lot of staff updated

not big but I saw a lot of lines updated

[ornate elk]

okay

then delete the venv folder

and relaunch

also wahts in your webui-user.bat?

[sweet basin]

works

getting this but it works

[ornate elk]

that doesnt look right xD

[sweet basin]

works tho

but getting errors on browser also

[ornate elk]

make sure its whitelisted in any adblocker

[tardy sparrow]

what is "easynegative" in negative prompt ?

do i need an extension to make it work ,

?

https://civitai.com/models/7808/easynegative?modelVersionId=9208

is it this one ?

[dreamy garnet]

it's an embedding/textual inversion, yes you need those models installed

[tardy sparrow]

so i just download it, put it in embeddings folder, use easynegative in negative prompt, then i'm done ?

[dreamy garnet]

use it in the "\stable-diffusion-webui\embeddings" folder

[tardy sparrow]

ya i did

what about badhands ? is it also an extension ?

[dreamy garnet]

after installing/restarting you use that in the negative prompt, as each one tells you in the readme (usually they have an info page or readme at civitai)

[tardy sparrow]

ya i read those

is "bad-hands-5" also an extension ?

[dreamy garnet]

yes i suppose so, just neg prompting that wont get you far

[tardy sparrow]

u mean neg promp is useless ?

[dreamy garnet]

in that case, probably, the "bad hands" embedding has been trained on bad hands, so you can prompt them out

but a neg prompt like "glasses" or "hat" is fine

[tardy sparrow]

mh

okay

thanks

can embeddings have the .pt extension ?

[ornate elk]

yes

[sour geyser]

Is console writing out "1006"\n an error code?

And if it is then what the heck does it even mean

[pearl musk]

Could it be that running a1111 without --medvram decreased my performance?

RX 6600XT, from 1.8 it/s with --medvram to 1.15it/s

[timber oak]

ayo

do u know what a 1006 error is?

img to img broken and txt to img fine

not for me but for @sour geyser

i tried googling it but it only shows "websocket error"

[stiff moat]

if i want to use multiple Loras how do I do that without it cursing all my generations.

[floral swallow]

hi guys any help? 😦

[timber oak]

its easy

you set their weight down

[stiff moat]

ive tried to but it wont come out clean

[timber oak]

sometimes they arent compatible

[stiff moat]

unless you tell me the lora is just trash

[hollow pike]

Some just don't work with certain models too

[timber oak]

imagine a lora with a person ALWAYS looking to the left

and then a lora where a persons head is always in a corner

[stiff moat]

Im using a character lora

[timber oak]

sometimes the ai will try to "fix" it

[stiff moat]

with some pose loras

[timber oak]

and make it look messy

in that case set the scale down

[stiff moat]

for which one?

[timber oak]

making your prompt weaker

cfg? or how its called

[stiff moat]

cfg is at 7

[timber oak]

the one u usually have between 6-8

[timber oak]

try 5

[stiff moat]

character is at .8

[timber oak]

same prompt

[stiff moat]

pose is .5

[timber oak]

yes sounds good

[timber oak]

maybe try .4

[hollow pike]

Try posting the error here. Does it happen with a different browser?

[timber oak]

it just says "1006" in the console

[stiff moat]

thanks ill give it a try and if nothing helps then maybe these loras dont mesh well together with eachother or the model

[timber oak]

when he starts img to img or inpaintit says:

stuff
1006
1006
1006

[timber oak]

yes

if the loras dont like each other just let the model do most of the work or separate the lorasd

OH

[timber oak]

you can try generating a character similar to you character in the lora

and then img to img

so like

if you wanna make princess zelda slashing some monster

just make it generate a blonde elf girl

then with inpaint or img to img

[stiff moat]

you mean just create my character first then img 2 img the other lora?

[timber oak]

you can try fixing it

[timber oak]

or that

[stiff moat]

i see

[timber oak]

depending if that works

[stiff moat]

thats actually not a bad idea

[timber oak]

id do that only if all other things dont work

SOMETIMES

the checkpoint does the funnies

[stiff moat]

yeah

its unavoidable

[timber oak]

so make sure to see if is better with other checkpoints

[stiff moat]

will do

[timber oak]

if it still doesnt work you can dm me if u want

but for now ill go to sleep

gn

[stiff moat]

thanks!

[vocal burrow]

incorrect python version

should be [3.10.6 ; 3.10.13]

once you've downgraded to a compatible version delete the venv folder

[floral swallow]

will do thanks

[slim gorge]

Hi, I'm trying to Train model in DreamBooth extention, but get error: Exception training model: 'The deprecation tuple ('LoRAAttnProcessor2_0', '0.26.0', 'Make sure use AttnProcessor2_0 instead by settingLoRA layers to self.{to_q,to_k,to_v,to_out[0]}.lora_layer respectively. This will be done automatically when using LoraLoaderMixin.load_lora_weights') should be removed since diffusers' version 0.26.1 is >= 0.26.0'. What I'm doing wrong?

[lament fog]

dont bother with the dreambooth extension

use kohya_ss

[slim gorge]

Mac user here...

[worn frigate]

Stable diffusion and controlnet both working now. thanks.
Wasn't possible without your help.

[slim gorge]

Trying to run kohya

Warning: LD_LIBRARY_PATH environment variable is not set.
Certain functionalities may not work correctly.
Please ensure that the required libraries are properly configured.

If you use WSL2 you may want to: export LD_LIBRARY_PATH=/usr/lib/wsl/lib/

[distant plover]

anyone know how to get sadtalker extension working in a1111?

[worn frigate]

is it supported?

[distant plover]

I see videos on using it in SD. I can't get it to work in SD or standalone...so who knows...

[worn frigate]

Are you able to install it?

[distant plover]

yes. It is installed, but it gives an error

[worn frigate]

whats the error

[worn frigate]

Hey @ornate elk so when using juggernaut xl with 25 sample steps and 1024 x 1024 with high res on.
No matter what i generate.
It finishes to 100% in the control panel but it never says or appears as final image on webpage or in output folder.

It shows this error to me

[distant plover]

Launching SadTalker Web UI
Traceback (most recent call last):
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\extensions\SadTalker\Launcher.py", line 204, in <module>
start()
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\extensions\SadTalker\Launcher.py", line 198, in start
demo = sadtalker_demo()
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\extensions\SadTalker\app_sadtalker.py", line 35, in sadtalker_demo
with gr.Row().style(equal_height=False):
AttributeError: 'Row' object has no attribute 'style'

Launch unsuccessful. Exiting.
Press any key to continue . . .

The stand alone won't run

[worn frigate]

I do not understand what it says but the last line ssaying cuda out of memory allocate more to pytorch. I had this error earlier too. Then i found out I have to add a line to web user bat and that will fix it.
This line in picture.

SO what do i do?
I remember you had told me this line is not necessary.
Does adding it make stable diffusion slower?

[distant plover]

this is the error in SD with the extension when I run it, try to generate:

using safetensor as default
{'checkpoint': 'C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker/checkpoints/SadTalker_V0.0.2_256.safetensors', 'dir_of_BFM_fitting': 'C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker/src/config', 'audio2pose_yaml_path': 'C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker/src/config\auido2pose.yaml', 'audio2exp_yaml_path': 'C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker/src/config\auido2exp.yaml', 'use_safetensor': True, 'mappingnet_checkpoint': 'C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker/checkpoints/mapping_00229-model.pth.tar', 'facerender_yaml': 'C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker/src/config\facerender.yaml'}
*** Error verifying pickled file from C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker/checkpoints/mapping_00229-model.pth.tar
*** The file may be malicious, so the program is not going to read it.
*** You can skip this check with --disable-safe-unpickle commandline argument.

---

Traceback (most recent call last):
  File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\modules\safe.py", line 137, in load_with_extra
    check_pt(filename, extra_handler)
  File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\modules\safe.py", line 83, in check_pt
    with zipfile.ZipFile(filename) as z:
  File "C:\Users\Frank\AppData\Local\Programs\Python\Python310\lib\zipfile.py", line 1249, in __init__
    self.fp = io.open(file, filemode)
FileNotFoundError: [Errno 2] No such file or directory: 'C:\\Users\\Frank\\Auto 1111\\stable-diffusion-webui/extensions/SadTalker/checkpoints/mapping_00229-model.pth.tar'

--

Traceback (most recent call last):
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\venv\lib\site-packages\gradio\routes.py", line 488, in run_predict
output = await app.get_blocks().process_api(
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\venv\lib\site-packages\gradio\blocks.py", line 1431, in process_api
result = await self.call_function(
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\venv\lib\site-packages\gradio\blocks.py", line 1103, in call_function
prediction = await anyio.to_thread.run_sync(
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\venv\lib\site-packages\anyio\to_thread.py", line 33, in run_sync return await get_asynclib().run_sync_in_worker_thread(
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\venv\lib\site-packages\anyio_backends_asyncio.py", line 877, in run_sync_in_worker_thread
return await future
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\venv\lib\site-packages\anyio_backends_asyncio.py", line 807, in run
result = context.run(func, *args)
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\venv\lib\site-packages\gradio\utils.py", line 707, in wrapper
response = f(*args, **kwargs)
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\modules\call_queue.py", line 13, in f
res = func(*args, **kwargs)
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker\src\gradio_demo.py", line 51, in test
self.animate_from_coeff = AnimateFromCoeff(self.sadtalker_paths, self.device)
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker\src\facerender\animate.py", line 70, in init
self.load_cpk_mapping(sadtalker_path['mappingnet_checkpoint'], mapping=mapping)
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker\src\facerender\animate.py", line 147, in load_cpk_mapping
mapping.load_state_dict(checkpoint['mapping'])
TypeError: 'NoneType' object is not subscriptable

[worn frigate]

Update,
Even after adding this line, it shows me same error.

I am changing it(max split size) to 512 now

[worn frigate]

Tried doing this and it does not work

Should I return back to my old UI?

Ok
So i set it to 1024 and it works but this is what results looks like

wait i forgot high res fix

I did everthing.

whats happening is.
it generates, to 100% works finee

then it auto deletes the image

this is so weirrrrd

It gets to total 100% generation and the image gets delete from everywhere

I am changiing it back to my old one.

[worn frigate]

changed back and its not working now.
In short stable diffusion is not working..

neither new webui text nor old one i used

[rotund lava]

add this next to the --api --xformers --medvram-sdxl it should look like this set COMMANDLINE_ARGS= --api --xformers --medvram-sdxl

[worn frigate]

whaaaaaaaaaaaaaat bro you are a scientist

it worked lmao

[distant plover]

cant see why this isn't running. Tried reinstalling everything, but still no luck

[worn frigate]

So i dont know whats wrong
but i am sending two pictures

this did not work

this worked.

[distant plover]

Traceback (most recent call last):
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\venv\lib\site-packages\gradio\routes.py", line 488, in run_predict
output = await app.get_blocks().process_api(
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\venv\lib\site-packages\gradio\blocks.py", line 1431, in process_api
result = await self.call_function(
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\venv\lib\site-packages\gradio\blocks.py", line 1103, in call_function
prediction = await anyio.to_thread.run_sync(
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\venv\lib\site-packages\anyio\to_thread.py", line 33, in run_sync
return await get_asynclib().run_sync_in_worker_thread(
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\venv\lib\site-packages\anyio_backends_asyncio.py", line 877, in run_sync_in_worker_thread
return await future
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\venv\lib\site-packages\anyio_backends_asyncio.py", line 807, in run
result = context.run(func, *args)
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\venv\lib\site-packages\gradio\utils.py", line 707, in wrapper
response = f(*args, **kwargs)
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\modules\call_queue.py", line 13, in f
res = func(*args, **kwargs)
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker\src\gradio_demo.py", line 51, in test
self.animate_from_coeff = AnimateFromCoeff(self.sadtalker_paths, self.device)
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker\src\facerender\animate.py", line 70, in init
self.load_cpk_mapping(sadtalker_path['mappingnet_checkpoint'], mapping=mapping)
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker\src\facerender\animate.py", line 147, in load_cpk_mapping
mapping.load_state_dict(checkpoint['mapping'])
TypeError: 'NoneType' object is not subscriptable

[worn frigate]

https://tenor.com/view/thanks-a-bunch-thank-you-very-much-thank-you-thankyou-ty-gif-8926252271357250819

[rotund lava]

yea 8gb nvidia gpus need that --xformers and --medvram-sdxl for sdxl models,the 12gb ones dont need the medvram thing just xformers

[distant plover]

doesn't launch SadTalker stand alone, error when run extension in a1111, extension won't activate in SDNext...installs, but doesn't appear as a tab and has a red dot

[violet inlet]

guys, there is a way to install triton on win 10?

[ornate elk]

With hires fix on?
That could be to much for your GPU
Also you have set hires steps to 0, another issue that can cause that error

[violet inlet]

the vae is the last and heaviest step

[distant plover]

no vae in sadtalker

[fair oxide]

Every single model I use, I get this error: "NansException: A tensor with all NaNs was produced in Unet. Use --disable-nan-check commandline argument to disable this check."
What to do?

[ornate elk]

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

also make sure the models you try are larger than 2gb

[fair oxide]

Radeon RX550

In the bat file:
@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --use-directml --lowvram --opt-sub-quad-attention --opt-split-attention-v1 --no-half-vae --upcast-sampling --no-half

call webui.bat

[distant plover]

been trying for hours to run this sadtalker extension or stand alone...no luck

[ornate elk]

Have you tried the dev version of sad talker?

[ornate elk]

That looks ok.
Have you checked the model file sizes?

[paper crag]

say theres something i want to check out. Novelai has this thing were if you were to drag in a picture created by it, the settings and prompt use for it can be read and copy those settings.
Does stablediffusion have a similar feature? where you can drag a picture and see exactly how it was made?

[ornate elk]

yes, its the PNG-Info tab

[paper crag]

nice, thanks

ah, one small question.

If the picture generated was afterwards modified with some other software (photoshop, sai, etc.). Will stable diffusion still be able to read it?

[ornate elk]

if photoshop overrides the meta data then no

[worn frigate]

Yes hires fix on. the issue was no vae line

the gpu handles everything after removing that.

Bro i swear to god this upscale thing does not work

Here are two images. No one can tell which one is upscaled. I tried all upscalers in Extras tab.

noticed it says its malecious

[ornate elk]

yea you can remove that file

[worn frigate]

what do i do

[ornate elk]

an use an other upscaler

[worn frigate]

none of them work

it outputs same image with same quality

all of these.

I compared 18 generated pictures side by side

they are all the same original image

[ornate elk]

thats strange

[vocal burrow]

throw the original picture and an upscaled one in there so we can compare

[worn frigate]

ye]

alright

ok when i tried this one it says this

when i try nearest it does not say anything but out puts same image which i am sending

upscaaled

original.

with swinr

with lsdr

i have firewall (tinywall) disabled and it has internet permission

i have been trying to upscale or atleast less noise this image for a hour now

and atlast used denoise in photoshop.

weird how SD is not doing it damn.

I am sending the request.py is says has error in 1291 line

[fair oxide]

Is there a way to search the models by size?

[ornate elk]

in the models/stable-diffusion folder it should show the sizes.
on civitai you cannot filter for that

[vocal burrow]

ok slow down... too much information for me to handle right now @worn frigate.
Let's go back to the basics, restart your sdwebui. Try to upscale your image by 2x using R-ESRGAN 2x+.
Then copy paste the whole console log in a .txt file and drop said file in there so we can see everything.

[worn frigate]

alright ty

launching webui
`@echo off

set PYTHON=C:\Users\Dipper\AppData\Local\Programs\Python\Python310\python.exe
set PYTORCH_CUDA_ALLOC_CONF=garbage_collection_threshold:0.6,max_split_size_mb:128
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --api --xformers --medvram-sdxl
git pull

call webui.bat`

went to extras tab

dropped image

i do not have this oneR-ESRGAN 2x+.

[fair oxide]

I meant in civitai, of course. If I can't filter by size, how am I supposed to find large models?

[vocal burrow]

(you probably don't want set PYTORCH_CUDA_ALLOC_CONF ..., chances are you're not smarter than the pytorch devteam when it comes to handling gargabe collection stuff. It's not gonna fix your upscaling problem tho)

[ornate elk]

every checkpoint is 2gb or larger. so filter for 1.5 models

[vocal burrow]

ok R-ESRGAN 4x+ then, resize by 4

[vocal burrow]

whole log please

[worn frigate]

its downloading

[vocal burrow]

there might be important stuff before that

[worn frigate]

not finished yet

[vocal burrow]

ok

FYI you should end up with something like that (converted the output to jpg because of discord limitation)

[worn frigate]

wrong sorry

damn

my clipboard

its 44mb

but yes its same now

[vocal burrow]

ok so it's working

[worn frigate]

yessss

will it work for other models too

[vocal burrow]

unless models are broken, yes it should

we didn't do anything tho apart from restarting your sdwebui.

Maybe it was already in a broken state when you started to upscale

[worn frigate]

hm ig because of firewall. when webui started i have firewall on which doesnt allow control panel to use internet

i later disabled it but it did not affect it unless we restarted

[vocal burrow]

maybe it blocked the upscale model download

[worn frigate]

ye

All working now. Thanks

[distant plover]

OK, 8+ hours...still can't get Sadtalker to work

[nocturne topaz]

standalone install or a1111 extension?

[distant plover]

either one

standalone won't run

in a1111 it errors during generation

[nocturne topaz]

ok.. starting with a1111 - whats the error message in the console?

[distant plover]

I will check

in the UI it just shows "ERROR"

In the cmd window, it shows the following

using safetensor as default
{'checkpoint': 'C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker/checkpoints/SadTalker_V0.0.2_256.safetensors', 'dir_of_BFM_fitting': 'C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker/src/config', 'audio2pose_yaml_path': 'C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker/src/config\auido2pose.yaml', 'audio2exp_yaml_path': 'C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker/src/config\auido2exp.yaml', 'use_safetensor': True, 'mappingnet_checkpoint': 'C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker/checkpoints/mapping_00229-model.pth.tar', 'facerender_yaml': 'C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker/src/config\facerender.yaml'}
*** Error verifying pickled file from C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker/checkpoints/mapping_00229-model.pth.tar
*** The file may be malicious, so the program is not going to read it.
*** You can skip this check with --disable-safe-unpickle commandline argument.

---

Traceback (most recent call last):
  File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\modules\safe.py", line 137, in load_with_extra
    check_pt(filename, extra_handler)
  File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\modules\safe.py", line 83, in check_pt
    with zipfile.ZipFile(filename) as z:
  File "C:\Users\Frank\AppData\Local\Programs\Python\Python310\lib\zipfile.py", line 1249, in __init__
    self.fp = io.open(file, filemode)
FileNotFoundError: [Errno 2] No such file or directory: 'C:\\Users\\Frank\\Auto 1111\\stable-diffusion-webui/extensions/SadTalker/checkpoints/mapping_00229-model.pth.tar'

Traceback (most recent call last):
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\venv\lib\site-packages\gradio\routes.py", line 488, in run_predict
output = await app.get_blocks().process_api(
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\venv\lib\site-packages\gradio\blocks.py", line 1431, in process_api
result = await self.call_function(
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\venv\lib\site-packages\gradio\blocks.py", line 1103, in call_function
prediction = await anyio.to_thread.run_sync(
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\venv\lib\site-packages\anyio\to_thread.py", line 33, in run_sync
return await get_asynclib().run_sync_in_worker_thread(
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\venv\lib\site-packages\anyio_backends_asyncio.py", line 877, in run_sync_in_worker_thread
return await future
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\venv\lib\site-packages\anyio_backends_asyncio.py", line 807, in run
result = context.run(func, *args)
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\venv\lib\site-packages\gradio\utils.py", line 707, in wrapper
response = f(*args, **kwargs)
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui\modules\call_queue.py", line 13, in f
res = func(*args, **kwargs)
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker\src\gradio_demo.py", line 51, in test
self.animate_from_coeff = AnimateFromCoeff(self.sadtalker_paths, self.device)
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker\src\facerender\animate.py", line 70, in init
self.load_cpk_mapping(sadtalker_path['mappingnet_checkpoint'], mapping=mapping)
File "C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker\src\facerender\animate.py", line 147, in load_cpk_mapping
mapping.load_state_dict(checkpoint['mapping'])
TypeError: 'NoneType' object is not subscriptable

[nocturne topaz]

you need to add the "--disable-safe-unpickle" to webui-user.bat file under commandline_args

[distant plover]

I do have the models in extensions/SadTalker/checkpoints

[nocturne topaz]

it's trying to load the model but it isn't allowed to

[distant plover]

they're safetensor files

[nocturne topaz]

and so because it hasn't loaded, its trying to use something that isn't there

[distant plover]

i am trying it...have to figure out where to add that

no idea where to add it

[nocturne topaz]

in the webui-user.bat file in your stable-diffusion-webui folder

[distant plover]

call webui.bat --disable-safe-unpickle

??

[nocturne topaz]

there will be a line that says COMMANDLINE_ARGS=

[distant plover]

OK, I added it there\

[nocturne topaz]

then shut down the command window and run webui-user.bat

[distant plover]

yes done

I saw it loaded with that command line

C:\Users\Frank\Auto 1111\stable-diffusion-webui\extensions\SadTalker\checkpoints is where the models are saved

[nocturne topaz]

awesome

[distant plover]

The 256 is named SadTalker_V0.0.2_256.safetensors, not that model.pt.tar name

awesome? It's still an error

[nocturne topaz]

ah wait.. hmm I'd suggest there are models missing.. I was saying awesome to the fact the first one loaded

[distant plover]

both models are in that dir

those are the only two models

the 256 and the 512

FileNotFoundError: [Errno 2] No such file or directory: 'C:\Users\Frank\Auto 1111\stable-diffusion-webui/extensions/SadTalker/checkpoints/mapping_00229-model.pth.tar'

Same error whether I choose 256 or 512 for face model resolution

[nocturne topaz]

id go and look in the settings inside the user interface, and clear the model directory, and then readd it - I think there is something funky going on as the slashes change direction which shouldn't happen

[ornate elk]

dont add this, as its a security risk

the problem is that @distant plover didnt followed the install instrucstions from here:
https://github.com/OpenTalker/SadTalker/blob/main/docs/webui_extension.md

you have to download 4 files

[distant plover]

I did follow them

step by step

[ornate elk]

and you made the folder?

checkpoints?

i tried just for fun to set it up now, and it worked perfectly

no errors

[nocturne topaz]

if using safetensors, risk is minimal

[distant plover]

I didn't realize I had to download those pth.,tar files too

Now I have them

i removed the pickle disable

[ornate elk]

it wasnt a safetensor:

[nocturne topaz]

fair call..

[distant plover]

I didnt have it at all

I thought the only files needed were the safetensors files

[ornate elk]

ah okay, yea all 4 files are needed

then it should work 🙂

[distant plover]

ok, trying a generation again

yes, that was it

thanks

[ornate elk]

np, to be fair in the instruction they dont say if all or only safetensor is needed

[distant plover]

With Sadtalker installed, at last I can have my revenge on the Jedi.

It will be Yoda himself who issues Order 66!

[ornate elk]

hahaha xD

feel free to share that xD

[distant plover]

"Order 66 execute shall you now!"

thank you both for the help

[tulip violet]

fam.... iam at the end of my energy and really need help... can someone here make happy...?

i am using linux

and followed all instructions

_faces_model, analyze_faces
File "/home/micha/ai/ComfyUI/custom_nodes/comfyui-reactor-node/scripts/reactor_faceswap.py", line 12, in <module>
from scripts.reactor_logger import logger
File "/home/micha/ai/ComfyUI/custom_nodes/comfyui-reactor-node/scripts/reactor_logger.py", line 6, in <module>
from reactor_utils import addLoggingLevel
File "/home/micha/ai/ComfyUI/custom_nodes/comfyui-reactor-node/reactor_utils.py", line 8, in <module>
from insightface.app.common import Face
ModuleNotFoundError: No module named 'insightface.app'

Cannot import /home/micha/ai/ComfyUI/custom_nodes/comfyui-reactor-node module for custom nodes: No module named 'insightface.app'

or is this the reason?

seems to be ok

[nocturne topaz]

youve restarted since doing the pip install insightface?

[tulip violet]

let me try

[unreal valley]

can anybody help me here, do i need to manually get the models for ipadapter in the correct directory or it installs with the models, cuz here I am missing the models

[tulip violet]

still the same error

[unreal valley]

this error I also had a day before, I saw that they mentioned a python update, the link was there on the github page

I just followed the instructions and it worked

[tulip violet]

so i need to update to 3.11 maybe?

[unreal valley]

I think its a minor update cuz they had separate links for 3.10 and 3.11

[ornate elk]

have you downloaded the prebuilt insightface module?

or did you installed the visual studio stuff for it?

[tulip violet]

i installed visual studio

on ubuntu linux

[ornate elk]

ahh okay

[tulip violet]

maybe i did not downloaded the module?

prebuilt

[tulip violet]

iam a little noob sry.. not sure what to do... update python?

[nocturne topaz]

download the new version from the website (unless you are on linux)

[tulip violet]

I. (For Windows users) If you still cannot build Insightface for some reasons or just don't want to install Visual Studio or VS C++ Build Tools - do the following:

(ComfyUI Portable) From the root folder check the version of Python:
run CMD and type python_embeded\python.exe -V
Download prebuilt Insightface package for Python 3.10 or for Python 3.11 (if in the previous step you see 3.11) or for Python 3.12 (if in the previous step you see 3.12) and put into the stable-diffusion-webui (A1111 or SD.Next) root folder (where you have "webui-user.bat" file) or into ComfyUI root folder if you use ComfyUI Portable
From the root folder run:
    (SD WebUI) CMD and .\venv\Scripts\activate
    (ComfyUI Portable) run CMD
Then update your PIP:
    (SD WebUI) python -m pip install -U pip
    (ComfyUI Portable) python_embeded\python.exe -m pip install -U pip
Then install Insightface:
    (SD WebUI) pip install insightface-0.7.3-cp310-cp310-win_amd64.whl (for 3.10) or pip install insightface-0.7.3-cp311-cp311-win_amd64.whl (for 3.11) or pip install insightface-0.7.3-cp312-cp312-win_amd64.whl (for 3.12)
    (ComfyUI Portable) python_embeded\python.exe -m pip install insightface-0.7.3-cp310-cp310-win_amd64.whl (for 3.10) or python_embeded\python.exe -m pip install insightface-0.7.3-cp311-cp311-win_amd64.whl (for 3.11) or python_embeded\python.exe -m pip install insightface-0.7.3-cp312-cp312-win_amd64.whl (for 3.12)

i think you mean this instructions

but iam linux user and have not the portable version

this instructions are descripe portable version how to install... how can i install if i have nor portable?

[tulip violet]

iam linux user

[south vigil]

Yooo, I really need help here
I cant start stable fusion sadly ( this is my second time opening it) I never used to open it after the first time so im a super noob, please help me its very overwhelming for me 😦
I opened it over a "copy" of the stable diffusion automatic
i really have 0 experience only followed a tutorial

[tulip violet]

[ReActor] - STATUS - Running v0.4.1-b9 in ComfyUI
Traceback (most recent call last):
File "/home/micha/ai/ComfyUI/nodes.py", line 1872, in load_custom_node
module_spec.loader.exec_module(module)
File "<frozen importlib._bootstrap_external>", line 883, in exec_module
File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed
File "/home/micha/ai/ComfyUI/custom_nodes/comfyui-reactor-node/init.py", line 22, in <module>
from .nodes import NODE_CLASS_MAPPINGS, NODE_DISPLAY_NAME_MAPPINGS
File "/home/micha/ai/ComfyUI/custom_nodes/comfyui-reactor-node/nodes.py", line 15, in <module>
from scripts.reactor_faceswap import FaceSwapScript, get_models, get_current_faces_model, analyze_faces
File "/home/micha/ai/ComfyUI/custom_nodes/comfyui-reactor-node/scripts/reactor_faceswap.py", line 12, in <module>
from scripts.reactor_logger import logger
File "/home/micha/ai/ComfyUI/custom_nodes/comfyui-reactor-node/scripts/reactor_logger.py", line 6, in <module>
from reactor_utils import addLoggingLevel
File "/home/micha/ai/ComfyUI/custom_nodes/comfyui-reactor-node/reactor_utils.py", line 8, in <module>
from insightface.app.common import Face
ModuleNotFoundError: No module named 'insightface.app'

Cannot import /home/micha/ai/ComfyUI/custom_nodes/comfyui-reactor-node module for custom nodes: No module named 'insightface.app'

noboy an idea? :/

nobody

[ornate elk]

you need to download them

[unreal valley]

could you guide me through this process, cuz I am a noob as well, just started a week ago

[ornate elk]

you need all of the .safetensor files from here:
https://huggingface.co/h94/IP-Adapter/tree/main/models
and them put these into the folder with the other ControlNet models

[rocky rune]

Hi, I'm not sure whether this belongs here, but here goes. I have Stable Diffusion XL set up and working, and I'm able to create great stuff with particular models. However, if I use the base model, I get something that's garbled. See the 3 images below. The first 2 are created with other models, the third with the base model. I'm using the web ui and I'm unclear on what I'm doing wrong. The prompt is this in all 3: a painting of a man in a space suit, a photorealistic painting by Donato Giancola, featured on cgsociety, space art, detailed painting, hyper realism, oil on canvas

[unreal valley]

not in the ipadapter models folder?

[ornate elk]

i dont know the exact comfyui folders

so you can put it there yes

[ornate elk]

hey, what vae file do you use?

[unreal valley]

any models for sdxl?

[ornate elk]

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

[unreal valley]

thanks

[rocky rune]

I have both of the following in the Stable-diffusion folder: sd_xl_base_1.0.safetensors and sd_xl_base_1.0_0.9vae.safetensors. I've tried both of those. In terms of the VAE, I wasn't initially clear on what this was, but after watching a video, I have this currently set in the Settings to this: vae-ft-mse-840000-ema-pruned.safetensors

[ornate elk]

thats the problem, the vae-ft-mse-84000 is a vae made for 1.5 based models.
sdxl models need a different vae. for example:
https://huggingface.co/stabilityai/sdxl-vae/tree/main

[rocky rune]

Awesome. It works. Thank you so much for your help @ornate elk

[ornate elk]

no problem 🙂

[zenith widget]

hey, i just wanted to ask if the time to generate something like this is normal? ive seen videos of people generating images at mach 5 and in this prompt its been 7 minutes and estimated 10 minutes left, some other prompts ive tried take 20-40mins

webui-user.bat
set COMMANDLINE_ARGS=--medvram --xformers --no-half

laptop specs
Intel(R) Core(TM) i5-10300H CPU @ 2.50GHz
NVIDIA GeForce GTX 1650 Ti

[ornate elk]

your using an sdxl model on a 4gb gpu

so yes thats normal

better use a 1.5 based model

that will be much faster

[pearl musk]

Is this okay? it/s are high but it takes 4 minutes on a 512x512+adetailer no upscale
40/40 [03:56<00:00, 3.91s/it]

[zenith widget]

much appreciated ill try that out ❤️

[pearl musk]

My bad, this is seconds per iteration, it's alright

[ornate elk]

whats your gpu?

[pearl musk]

on this laptop, a GTX 950M 4GB

I have a 6600XT on my main rig, soon to be replaced with a P40 24GB

[ornate elk]

for the laptop make sure to use --xformers --medvram

[pearl musk]

xformers is not showing on the Optimizations tab on A1111

cmdline= Launching Web UI with arguments: --listen --medvram --xformers --opt-sub-quad-attention --always-batch-cond-uncond --upcast-sampling --device-id=0 --enable-insecure-extension-access

[ornate elk]

you have to check the bottom of the webui in browser

[pearl musk]

Oh, it is showing

[ornate elk]

also you can remove --always-batch-cond-uncond --opt-sub-quad-attention

[pearl musk]

Good, let me check

[ornate elk]

and test if its faster

[pearl musk]

Should I leave Cross attention optimization on Automatic?

[ornate elk]

on automatic, yes

it will use xformers

[pearl musk]

Applying attention optimization: Doggettx... done

is that okay?

[ornate elk]

no it should say xformers

[ornate elk]

you may need to install cuda toolkit 11.8

[pearl musk]

Okay, let me try that

[vocal burrow]

You shouldn't have to install anything additionaly, weird. What does the console log says after launch ?

Also what's your hardware ?

[pearl musk]

Using TCMalloc: libtcmalloc_minimal.so.4 Python 3.11.6 (main, Nov 14 2023, 09:36:21) [GCC 13.2.1 20230801] Version: v1.7.0 Commit hash: cf2772fab0af5573da775e7437e6acdca424f26e Faceswaplab : Use GPU requirements Checking faceswaplab requirements 0.00862452100000155 Launching Web UI with arguments: --listen --medvram --xformers --upcast-sampling --device-id=0 --enable-insecure-extension-access Tag Autocomplete: Could not locate model-keyword extension, Lora trigger word completion will be limited to those added through the extra networks menu. [-] ADetailer initialized. version: 24.1.2, num models: 14 ControlNet preprocessor location: /home/void/stable-diffusion-webui/extensions/sd-webui-controlnet/annotator/downloads 2024-02-08 11:22:21,058 - ControlNet - INFO - ControlNet v1.1.440 2024-02-08 11:22:21,227 - ControlNet - INFO - ControlNet v1.1.440 Loading weights [6214e8dd70] from /home/void/stable-diffusion-webui/models/Stable-diffusion/kizukiAnimeHentai_animeHentaiV3.safetensors 2024-02-08 11:22:22,027 - AnimateDiff - INFO - Injecting LCM to UI. /home/void/stable-diffusion-webui/modules/gradio_extensons.py:25: GradioDeprecationWarning: optional parameter is deprecated, and it has no effect res = original_IOComponent_init(self, *args, **kwargs) 2024-02-08 11:22:22,769 - AnimateDiff - INFO - Hacking i2i-batch. 2024-02-08 11:22:22,842 - ControlNet - INFO - ControlNet UI callback registered. Creating model from config: /home/void/stable-diffusion-webui/configs/v1-inference.yaml Running on local URL: http://0.0.0.0:7860

[ornate elk]

wrong python version

SD needs a version between 3.10.6 up to 3.10.11

[pearl musk]

crap, I'm on Arch, how do create a 3.10.6 venv?

[pearl musk]

GTX 950M, i7 6700hq, 12GB DDR4

maybe xformers is not available for Maxwell GPUs anymore?

[ornate elk]

thats my second guess, but then you can build it yourself

[vocal burrow]

Theoretically it can work with 3.11 now but it's probably best to stick to what's known to be working.

[pearl musk]

I will try to make a 3.10.6 venv and try again

[ornate elk]

https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Xformers#building-xformers-on-linux-from-anonymous-user

[vocal burrow]

what happens if you set --reinstall-xformers ?

[night orchid]

what can I do with this ? " OutOfMemoryError: CUDA out of memory. Tried to allocate 50.00 MiB (GPU 0; 12.00 GiB total capacity; 10.12 GiB already allocated; 0 bytes free; 10.81 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF"

[ornate elk]

i think thats why it didnt get installed properly

[ornate elk]

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

[night orchid]

my GPU is Nvidia RTX 3060 12GB

set COMMANDLINE_ARGS= --xformers --no-half --disable-nan-check

[pearl musk]

Nothing changed with --reinstall-xformers

[vocal burrow]

no hint of the sdwebui trying to get / install anything ?

if so I'd try to build xformers like cs1o mentioned

[pearl musk]

Using TCMalloc: libtcmalloc_minimal.so.4 Python 3.11.6 (main, Nov 14 2023, 09:36:21) [GCC 13.2.1 20230801] Version: v1.7.0 Commit hash: cf2772fab0af5573da775e7437e6acdca424f26e Installing xformers Faceswaplab : Use GPU requirements Checking faceswaplab requirements 0.009282667000661604 Launching Web UI with arguments: --reinstall-xformers --no-gradio-queue --listen --medvram --xformers --upcast-sampling --device-id=0 --enable-insecure-extension-access

[vocal burrow]

it would probably be under that

oh

wait

Installing xformers