🤝｜tech-support

[rugged latch]

aw, so i have to resinsatall stable completky

[crystal nacelle]

maybe. he said you can start at step 5 bru

so maybe all you have to do is add some command lines...try that first

[rugged latch]

did

and it said nu

so i get to reinstall it

:,3

where is stable's CSS script located

i dunt rememebr

wut do i do now

cant start it, and followed instructions

[rugged latch]

oki, bird had big brain and fixed it, everything working now

even mai custum css and etc

thx chuu

dunno why task manager is saying 0% gpu usage, but the "in-use" bar is at max usage

prolly an issue wit windows or sum, buh thx uu

[rugged latch]

didnt expect this, GPU renders stuff diffrent;y than CPU does, sum how GPU renders looks worse but faster <v>

[mystic basin]

is it normal that it doesn't find my lora?

i've got two lora

[crystal nacelle]

refresh it

[mystic basin]

i tried

but doesn't work

[crystal nacelle]

go to extensions

tick everything (especially lora) then apply andd restart UI

maybe check fr updates first then apply and restart

[mystic basin]

alright i'll try

nope

doesn't work

[rugged latch]

question

why does CPU only mode follow prompts better than GPU

<v>

[mystic basin]

i don't know

[rugged latch]

is that a thing that happens ovO

or is this unheard of

[fair oxide]

im getting this error code shortly after starting lora training on kohya

[rugged latch]

another question >:3

is there a way to see the image generation happen live

as its happing

[rugged latch]

cant really see it ;v;

is blurry

[pearl musk]

Every time I load a1111, if I leave a VAE loaded from a prior session, VAEs bug and I have to reload them, is this normal?

[fair oxide]

sorry this should be a better picture

[rugged latch]

to me it looks like wut ever ur trying to connect to on huggingface is down

atm

maybe try again later ovO

Ya is def the server :^

code 500 is thrown when the server dunno wuts happening

[fair oxide]

i think i found a solution :>

[mystic basin]

why doesn't any of my lora show?

i have three lora

[rugged latch]

wut in the world are chu using <v>

dunt look like any stable i ever saw

[fair oxide]

atleast this is looking allot better than what was going on before

[rugged latch]

aye

chu fix it :3

[fair oxide]

:3

[rugged latch]

wut ended up being the issue, so others who have the issue can look :33

[fair oxide]

well it was the model that i was training the lora on, i just tried a few different ones and this one i picked now ended up working (atleast so far) :>

[rugged latch]

noice :3

[fair oxide]

indeed :3

[mystic basin]

could this be because i use the amd version?

[rugged latch]

i dunt think so

bc i am usig AMD verison, infact i have two verions that are AMD, and i never have any issues with loras or models

try this

type --disable-safe-unpickle and --enable-insecure-extension-access

in the command args

[mystic basin]

alright thx i'll try

still doesn't work

i don't know if it's because of any of this

[fair oxide]

new error :(

and its working again ?????

this is the most confusing shit ive ever seen 😭

eh i give up 😭

[somber flame]

it's going to be hard doing this if you are taking screenshots of the terminal and then missing the leftmost column when you do that

you can try the UIs for lora training. there is nothing in particular special about kohya's scripts

[somber flame]

you didn't put in the flags as the user told you to, but i don't think that has anything to do with your issue. try comfyui

[civic hare]

can someone help me out with what bitsandbytes is? :3

[civic hare]

(these are my screenshots btw :3 were working together)

[nocturne topaz]

#🔧｜finetune

[worn frigate]

help pls

[frosty shuttle]

help i typed
dog
in the prompt

this is what came out

[worn frigate]

damn lel

[frosty shuttle]

wtf is wrong

with my thing

every prompt i put

[worn frigate]

what model are u using

[frosty shuttle]

the result has nothing to do with it

[worn frigate]

sameee

[frosty shuttle]

itsss

[worn frigate]

bruh sane

same

look at my problem above

its same for me

[frosty shuttle]

SDXL

[worn frigate]

damn bruh whats this

[frosty shuttle]

brooo wtfff

yeah i had that too earlier

but like

its getting worse

and worse

it was generating decent at the start but now iwth horrible

[worn frigate]

same i am gettin abominations now

[frosty shuttle]

😭

im gonna tryrestarting my pc

[somber flame]

what is your CFG? are you using... what?

there's not enough information

[frosty shuttle]

dude i just started this

idk shit but i did 3 hours worth of installation

[somber flame]

well

[frosty shuttle]

what is CFG

config?

[somber flame]

no it stands for classifier free guidance

[frosty shuttle]

how do i know

[somber flame]

lol

[frosty shuttle]

how do i know my CFG?

this my ting

[worn frigate]

ots jere

belong height

CFG scale

[frosty shuttle]

ohhhh

kk

I should change it?

[worn frigate]

any idea how to fix this

[worn frigate]

nah

[frosty shuttle]

also my pics are taking awfully long to generated compared to earlier

[worn frigate]

restart stable diffusin

[frosty shuttle]

i did

i restarted my whole pc

i put x former

and now its slow af

[worn frigate]

do u have nvidia gpu

[frosty shuttle]

yes

[worn frigate]

how much ram

[frosty shuttle]

great now its creating absolute horror

hmm

let me check

yeah

i dont think i have enough

[frosty shuttle]

16 GB

[worn frigate]

then add

uh nothing

because xformer helps

send full webui text

[frosty shuttle]

i added something else

that might be it

[worn frigate]

which model are you using though

[frosty shuttle]

sdxl

[worn frigate]

wait i think i got it

[frosty shuttle]

base

[worn frigate]

base model which comes with stable diffusion

u have only 1 ?

[frosty shuttle]

yep

[worn frigate]

bro that model is trash

[worn frigate]

it never generated anything good lmao

nah man its a waaste

[frosty shuttle]

idk they have a whole

category

for it in the discord server

seemed like something

[worn frigate]

imma send u one

[frosty shuttle]

#✨｜sdxl

kk thanks

[worn frigate]

ok before that

i want you to set the resolution higher in case itss 512

set it to 1024

[frosty shuttle]

ok ok

[worn frigate]

make it 1024

then run command

[frosty shuttle]

im reopening the thing rn

ok kk k

i removed things except xformer

[frosty shuttle]

here

[worn frigate]

you dont need medvram because u higher than 12 gb

[frosty shuttle]

should i put refiner?

[frosty shuttle]

ahh kk

its saying this

this looks bad

failed to create a model 😭

is it not even seeing the current one?

[worn frigate]

its downloading

[frosty shuttle]

ahh kk

[worn frigate]

meanwhile download one good model

[frosty shuttle]

which model should i download

[worn frigate]

https://civitai.com/models/133005/juggernaut-xl

get this one

its XL

[frosty shuttle]

where download

nevermind

im dumb af

its downloading

there is 0 space left on my C drive

but i installed everything on D

wtf is going on

[worn frigate]

show cmd

[frosty shuttle]

cleared

[maiden haven]

im not sure if ive had one error yet, definitely nothing package related

[frosty shuttle]

@worn frigate i dont get it

earlier it was taking

2 minutes

now its been 5 minutes

[worn frigate]

its download like 10.2 GB file

no wonder why you are running out of space

[frosty shuttle]

and its probably just gonna whip out the most eye burning image

i found the cause of disk usage on my C drive

i got like 100 left in my d

and i got 12 in my C

[worn frigate]

Its a sdxl model.

[frosty shuttle]

mf's forgot to download the important stuff an hour ago 😭

[worn frigate]

So it is installing stuff it needs

[frosty shuttle]

ahhh okk

[worn frigate]

We cant do anything about it

[frosty shuttle]

how much gb is gonna be left after like a week of usage on this

[worn frigate]

the cmd says 1.7 and 10 so 11.7 total yet

not sure if it will downloading anything else

[worn frigate]

working now?

[frosty shuttle]

hoild on

omg

yes

wtf

[worn frigate]

lol

[frosty shuttle]

this is god

this

omg

finally

[worn frigate]

YEE

the model working now?

or its the downloaded one

[frosty shuttle]

this the smae model

[worn frigate]

les go

[frosty shuttle]

the downloaded one just needs to be setuop

[worn frigate]

good

[rugged latch]

boop

is there a way to watch stuff get generated

like, a live preview

[frosty shuttle]

yeah its blurry for me

im using ui thing

web ui

[maiden haven]

comfyui has a nice live preview, i havent used it much yet but sd-webuis was pretty bad

im sure some extensions better

[rugged latch]

how u get it where u can watch it ;^;

I was see it happen

[maiden haven]

sd-web-ui had it too, it has a big skip/interrupt button

i have installed so many cuda nvidia things today

120am

it was an absolutely smart idea to go to wsl2 though im so much less aggrevated, its just taking awhile

round like.. 15

[crystal nacelle]

Can i get AnimateDiff to work on an AMD GPU with 8GB ram? I keep getting the not enough memory error

[nocturne topaz]

Probably not unless doing 256x256

[crystal nacelle]

know any free online AI vid tools? i can atleast generate an image with stable diffusion then use online AI for the video?

[ornate elk]

You just have to switch your model to see the loras.
Remove the --disable in any check --insecure extension access.

Loras are either for 1.5 or sdxl, so the only show up when selecting the right base model

[ornate elk]

Normaly there is image preview enabled in the settings

Also GPU doesn't generate images with worse quality.

[ornate elk]

It requires to much vram, I dont think so

[tribal thunder]

how to generate images

[sweet basin]

what to do when you reinstalled a windows

[vocal burrow]

you didin't install git

[sweet basin]

I kept the files so its the same sd laucher

damn how to install it now

been ages since I installed sd

nvm

thought its a bit complicated process

installed git and it works now

[heady pine]

Is there a way to find out what image you used as a base on an open pose image you made? I made a great looking image with an open pose image, but have forgoten what image was used

[vocal burrow]

probably not

[mystic basin]

it does work now when i change of model

thanks for the help

[fiery pumice]

I've got this error, but even after entering and saving what I put in, it still tells me I need to disable the check even though I already have, everywhere I looked online I couldn't find a solution.
I'm told to put in the COMMANDLINE_ARGS= --skip-torch-cuda-test when I've already done so and I'm still stuck at square one. Any options of what may be happening? I did test by just having --skip-torch-cuda-test but that didn't work.

[maiden sonnet]

I was trying to deploy and run Stable Diffusion Webui in an HPC Cluster. But the application is giving http://127.0.0.1:7860 as an output IP host, which I can't visualize inside the HPC cluster.

I've also tried to perform SSH Tunnelling to my local machine, but no luck.

$ ssh -L 8080:127.0.0.1:7860 proddev@10.172.X.X

Appreciate any help.

[wind sun]

face fusion

not working

please help

[pearl musk]

Has anyone used SAG (self attention guidance)?

[ornate ruin]

Hi:) on launch I always get the 'style data base not foud' message. Is that something I can "fix"? To speed up the launch time maybe?

[vocal burrow]

what's your hardware ?

[vocal burrow]

Use set COMMANDLINE_ARGS=--share --gradio-auth xxx:yyy where xxx is your desired login and yyy your password

[vocal burrow]

It just means you haven't saved any style for your prompt. It's ok.

[ornate ruin]

Okay thanks :)

[maiden sonnet]

forgive my ignorance, but do I need to put this insider webui-user.sh file?

[vocal burrow]

Yes there should be a line for you to uncomment.

If you're using nvidia hardware, you might as well throw --xformers in there.

Also I'd edit that 10.172.x.y ip if I were you.

lol nevermind, I'm tired

[maiden haven]

https://tenor.com/view/spongebob-gif-12803876568958189142

i think i can get everything up today

[maiden sonnet]

you are god sent. it worked

[ornate elk]

Make sure you close the webui if you try to run FaceFusion, as they use the same port

[ornate ruin]

Is a cuda run time error normal for upscaling the image too large in the text to image tab? Or is that something I should to something about.?

Tried to upscale a 512x768 image by x4 with esrgan

30 steps and 15 hires steps

[vocal burrow]

depends of your hardware

[ornate ruin]

So it's likely normal. Okay. Because when that happened before it didn't show that but just said out of memory underneath where the image should have generated

[ornate ruin]

That's what it said btw.

RTX4060 mobile
64gb ram
I7-13700k
8gb vram

[vocal burrow]

Use set COMMANDLINE_ARGS=--xformers --no-half-vae --medvram-sdxl in your webui-user.bat

[ornate ruin]

I run: --xformers --medvram-sdxl --no-half-vae

[ornate ruin]

Oh I wrote that too late, sorry .lol. yes I already use them

[maiden haven]

win + shift + S opens snippet for screenshots

[ornate ruin]

If I may ask another thing .. so now I have gotten an image I like, 512x786 and I loaded it into the img2img tab to upscale it -

Empty prompts just 'highly detailed' for positives
Just Resize tab:
Set the width and height parameters, chose Script SD Upscale with the 4x ultrashar, scale factor 2 and tile overlap unchanged at 64.
36 sampling steps, 0.1 denoise,

First generation did seem fine, couldn't see much change but it was fairly fast at under 5 minutes.

I reloaded the new image into the img2img tab and hit generation again and now it takes over an hour for the second run..

Is that normal?

[maiden sonnet]

What's seems to be wrong with the output images I am getting? I'm getting output images like the attached from different models that I am using. I am using the base models available from huggingface, and CIVAT models.

[maiden haven]

i spent zero time on it but i had that happen with a couple of models but hte rest of mine work fine not sure whats up

im at 40 minutes lol => => # Installing collected packages: mpmath, typing-extensions, sympy, pillow, nvidia-nvtx-cu12, nvidia-nvjitlink-cu12, nvidia-nccl-cu12, nvidia-curand => => # -cu12, nvidia-cufft-cu12, nvidia-cuda-runtime-cu12, nvidia-cuda-nvrtc-cu12, nvidia-cuda-cupti-cu12, nvidia-cublas-cu12, numpy, networkx, MarkupSa => => # fe, fsspec, filelock, triton, nvidia-cusparse-cu12, nvidia-cudnn-cu12, jinja2, nvidia-cusolver-cu12, torch, torchvision, torchaudio

[vocal burrow]

what settings are you using ? screenshot your webpage @ornate ruin @maiden sonnet

[nova ermine]

I'm using lineart_realistic preprocessior in controlNet

I get this map and then disable preprocessor

[vocal burrow]

resizing to 2048x3072, that's crazy high.

[ornate ruin]

excuse the redline, accident

[nova ermine]

But then when I use ControlNet, my image gets a lot of artifacts on the edges. How to fix?

[vocal burrow]

If you want to upscale to such absurd resolution, try https://github.com/Coyote-A/ultimate-upscale-for-automatic1111

How does Ultimate Upscale Script works.
Basically it's gonna split your (too) big image into multiple smaller ones called tiles. Upscale each one individually and then stitch them back together to make the higher resolution output.
What do those settings do ? cf screenshot
why blur the edges ? Because it helps them blend better. SD will figure out something correct from the blurry mess, don't worry.
https://cdn.discordapp.com/attachments/1091193032273043516/1194628488200724550/image.png?ex=65ccbafa&is=65ba45fa&hm=135890c7d54e5672337696c1aa7dc4b6c6efae312a9332da5fba38ed729f8b9c&

[ornate ruin]

oh? but how do i continue upscaling an image then? do i put 512x768 in there?

[vocal burrow]

tbf there s a lot of artifacts on the preprocessed image.

[maiden sonnet]

Most of the time, I get just like this - gray image.

[vocal burrow]

what's your hardware and what does the console says ?

[maiden sonnet]

NVIDIA A100 80GB,

[nova ermine]

Yeah, but how to avoid getting that artifacts on the proprocesed image (the following image is the one that I use to generate the preprocessed image)

[vocal burrow]

are you using a lora as a model ?

[maiden sonnet]

No

[vocal burrow]

what's this model-lora-food-10.safetensors then ?

[vocal burrow]

in there where ? you set the desired resolution in the "Resize to" section just like you did.

[maiden sonnet]

these are all my models. I just wacked it all in under the models/Stable-diffusion folder

[vocal burrow]

what's the size of that model ?

Just from the name I'm guessing that "burger photo", "model-lora-food", "foodstyle", "yummyfood" are loras

[maiden sonnet]

just 18MB

[vocal burrow]

yup that's not a model

models are at least 2gb

those are loras (or less likely embeddings)

[maiden haven]

ahh there we go, broke wsl2.. add 2 hours on

[maiden sonnet]

ah well now that makes more sense. My question I guess, is how I can use these Lora models for e.g for txt2img?

[ornate ruin]

Tbf I'm not familiar. I just read in an article that that's a way to indefinitely upscale an image (upscale the way I did, then repeat)

With in there I mean as the set resolution..but I don't think that'd make any sense to get a higher resolution image.

[vocal burrow]

LoRa/Lycoris quick guide :
1/ Where do I put my LoRa : stable-diffusion-webui\models\Lora, Loras are NOT stable-diffusion model but rather patches for said models.
2/ How do I use my LoRa, you can either :

• Invoke it directly in your prompt by using the following syntax <lora:filename:weight> (or <lora:alias:weight> depending of the setting in setting->extra network). eg : <lora:animeTarotCardArtStyleLora_v20Offset:1> for using the Lora animeTarotCardArtStyleLora_v20Offset.safetensors
• Use the UI "Extra Networks" / card system by doing as described in my screenshot. Clicking an associated lora card should automatically insert the lora prompt liked mentioned above. https://cdn.discordapp.com/attachments/1002602742667280404/1191779742534348870/loraHowTo17.jpg
  3/ Be careful, LoRa are trained specifically for some SD version. Lora trained for SD1.5 (and derivatives) models will not load on SD2 models and vice versa (same goes for SDXL). Also even though Lora trained on SD1.5 models should work for every SD1.5 derivatives, they'll probably work best with the model that has been used for its training. Therefore, (by default) only Loras compatible with the model you currently have loaded will show up.
  4/ LoRa, LoHa, LoCon, what are these ? They're all commonly refered as "Lora" but differ slightly technically :
• LoHa : Lora with hadamard decomposition
• LoCon : Lora extended to redisual blocks.
• More explanations : https://stable-diffusion-art.com/lora/, https://www.sdcompendium.com/doku.php?id=lora_0014, https://www.sdcompendium.com/doku.php?id=locon_training_0071, etc

[vocal burrow]

There are many many ways to upscale an image. Each with their up and downs. But an "infinite" upscaler do not exist, at some point you'll add noises, artifacts, consume too much vram, etc.

Try to use ultimate sd upscaler

[ornate ruin]

For reference that's what I've tried to replicate

https://www.reddit.com/r/StableDiffusion/s/xO1s8FdbX8

[vocal burrow]

There's not much you can tweak on the preprocessors afaik, you can try a different preprocessor model, blur the edges of the input image, etc

[maiden sonnet]

Thanks. will go through this

[vocal burrow]

Wow I haven't seen that tutorial in ages. That's basically prehistoric.

[ornate ruin]

Oh 😅

[vocal burrow]

It dates back probably to the first or second month of the initial SD release

it's not wrong but there are better ways to do so now.

[ornate ruin]

That means technically that it should be even easier now days tho, no?

[vocal burrow]

not necessarily

I'd advice you again to use ultimate-sd upscaler for a first upscaling.

[ornate ruin]

I also don't need absurd resolutions. Just enough to be able to zoom in without seeing artifacts or pixles

[vocal burrow]

it's probably one of the easier and robust way to upscale

[ornate ruin]

Why is that?

Ah sorry I read 'against using it's my bad. I'll look into a tutorial for that

[vocal burrow]

Because it does tiling better than the built-in method, it gives you more control.

[vocal burrow]

I've already sent you one #🤝｜tech-support message

Basically, set the desired resolution in "Resize to" section, tweak the cfg, denoiser (or let it as is for now) and input some generic positive and negative prompts that apply to your overall image

[ornate ruin]

Do I just copy the link and paste it within automatic1111 in the add extensions tab?

[vocal burrow]

Important to have a generic prompt as that prompt is gonna apply to each tile of your image

[vocal burrow]

it should be in the list of available extensions

click "load from"

then search for it

[ornate ruin]

Preparing dinner right now. I'll try it out later and if I have troubles I hope it's okay if I @ you again

[vocal burrow]

I might be busy but try it

[ornate ruin]

Don't worry . Thanks for your help 🙂

[nova ermine]

Do you mean, bluring the edges in Photoshop?

[quiet swift]

Hello, I try to create one image in img2img, but SD tells me : RuntimeError: MPS backend out of memory (MPS allocated: 34.46 GB, other allocations: 381.20 MB, max allowed: 36.27 GB). Tried to allocate 3.96 GB on private pool. Use PYTORCH_MPS_HIGH_WATERMARK_RATIO=0.0 to disable upper limit for memory allocations (may cause system failure). Do you know what I can do to solve this trouble ? (I use a Mac.) Thank you in advance if someone can help.

[nocturne topaz]

On a mac, the options are to close other programs (especially browser tabs), try a smaller resolution, or do a google search for the command that allows you to allocate more memory to the gpu (https://www.reddit.com/r/LocalLLaMA/comments/186phti/m1m2m3_increase_vram_allocation_with_sudo_sysctl/)

[quiet swift]

thank you for your reply.

[vocal burrow]

yes (or gimp, or paint.net or whatever)

[nova ermine]

I've got the problem

I was using an image with transparent bg, and for any reason SD doesn't get on well with that

So I've got this image (the same but with sky bg)

Look at the differences on the maps

[wind sun]

still no video for output

[quiet swift]

I have checked your link. Ouahhh, it's way too much complex for my little brain 🙂 I will have to go in RunDiffusion 😦 (I have closed my other applications, closed most of the window of my firefox browser, reduce the size of my image but that did not change the trouble)

[wind sun]

@karmic crown

[nocturne topaz]

it's pretty easy - hit command space and type terminal and press enter - then in the window that pops up type sudo sysctl iogpu.wired_limit_mb=40960 then enter your password and done

[quiet swift]

I did not thought it was so easy. I will try that. Thank you. Nevertheless, they say "Will this make your system less stable? Probably." 🙂

[nocturne topaz]

the only reason it becomes unstable is because people get too aggressive allocating most of the ram to the gpu, if you are sensible and leave a good amount for the system, you will be fine

[quiet swift]

I've got a last question : if I do what you say, and then, later, I want my mac to be as before (sorry for my poor english, I'm french), is it easy to have the VRAM as before ? Maybe, if I just turn off my computer ?

[nocturne topaz]

hence why I put 40960 which should leave 8gb for the system

[nocturne topaz]

Restarting the mac should reset it back to normal

[quiet swift]

ok thank you for your explanations 🙂

[nocturne topaz]

I'm guessing from the fact it oomed at 36gb that you have a 2023 mbp with 48gb of memory

if you only have 36gb, then youll want to set the limit to 28762

[sonic ferry]

Hey, how do I get a preview on Ultimate SD Upscale?

[nocturne topaz]

I'm guessing you are using comfy, as a1111 the previews tend to come through as normal?

In which case, not sure...

[sonic ferry]

Oh yes, this is ComfyUI

I've seen screenshots where there is a preview on the bottom same as the KSamplers get as it's processing but I'm not seeing anything on the Upscaler

[ornate ruin]

clicking load from just shows errors everywhere

[wind sun]

can somebody help me

[celest talon]

Hi everyone, does anybody know what's the best option/model for image faceswapping?

[valid summit]

can someone help ?

my stable diffusion dont select the checkpoint, it loading the checkpoint but the box stills empty like this

and here shows this error message

and in cmd, before sd starts, the "no modules xformers processing without it" shows too

[ornate elk]

Hey, what's your GPU?

[valid summit]

GTX 1650

4GB

i have already used sd before

on this gpu

[ornate elk]

Okay, make sure the webui is whitelisted in any browser adblocker

[valid summit]

ok, lemme check

my adblock has disabled

u think if i add --xformers and --medvram commands will fix this ?

[ornate elk]

You need these for sure

But I dont think they fix that

[valid summit]

i'll test

lol, now the sd dont start and shows another message on cmd

[ornate elk]

What error

[valid summit]

wait im accidentally close the cmd, im open again

lol its take too long in this, its normal?

[rugged latch]

idk wym, thats art

wut are u seeing

[valid summit]

@ornate elk i have been added the commands and the stable diffusion stills with the same error

[ornate elk]

Can you try an other browser?

[valid summit]

yes i can, wait

when i try to change the checkpoint, this appers on cmd

here are some error when that i found

and this error repeat themselves with all the cmd

[nocturne topaz]

That's usually a dud download

[valid summit]

bad download?

like corrupted files?

[valid summit]

its normal get this in stable diffusion instalation?

ok is not normal, how i can fix this??

[ornate elk]

You need a stable internet connection. Best with Ethernet.
Then make sure you dont have any third party antivirus Programm running.
You also may need to change your DNS Server or use a VPN

[valid summit]

i'll try vpn

oh i fixed

i just have synced the windows clock

[steady thistle]

For some reason I have less vram available now after updating controlnet and certain features. Is there something that can be done about the whole onnx call thing?

[ornate elk]

Make sure your using the 700mb controlnet models

[steady thistle]

I didn't realize that was an option. those are the .bin files?

[ornate elk]

Depends on the controlnet you want to use

For 1.5 there are 700mb models

[steady thistle]

I'm not sure where to look actually. Hmm, ok.

[ornate elk]

1.5 controlnet models:
https://civitai.com/models/38784/controlnet-11-models

Sdxl controlnet models:
https://huggingface.co/lllyasviel/sd_control_collection/tree/main

[ornate elk]

The error you get is triggered by an extension maybe. Also the "Please restart server" is from Reactor. Seems like its broken so a reinstall of Reactor is an idea

[steady thistle]

Ah, okay! Thanks, I couldn't tell it was from ReActor.

[valid summit]

BRO I HAVE RE-INSTALLED THIS SHIT BUT STILLS WITH THE SAME PROBLEM, PLS SOMEONE HELP

[wind sun]

yo what about me

yall forgot me

[vocal burrow]

what does the console says ?

[vocal burrow]

what's the full log ?

[valid summit]

how i can send the full log?

[vocal burrow]

copy paste the whole output of the console instead of screenshoting parts of it.

(put it in a .txt and drop if in there for ease of read)

[valid summit]

ok

i cant send here anywaws, have 35k characters

[vocal burrow]

what's your hardware ?

[valid summit]

I5 9300h GTX 1650 4GB 16GB RAM

i've tryed with --medvram but stills the same error

[vocal burrow]

use --autolaunch --lowvram-sdxl --medvram --xformers

there's no reason to use --disable-safe-unpickle

I wouldn't use --skip-version-check either

then, what's the size of 7pagNSFWMergedModel_7pagLSCX.safetensors ?

[valid summit]

its a checkpoint file

oh wait

i read wrong

its 3.97GB

[valid summit]

i got this with this arguments

[vocal burrow]

check that its SHA256 is indeed DDA4C6CFE3A38B58366C1B76FF9A693E50D71DD37174F916156FB5C778C84F7C

[vocal burrow]

mabad, remove --lowvram-sdxl

[valid summit]

and how do I check this?

[vocal burrow]

if you have 7zip, right click the file, 7-zip->CRC SHA->SHA-256

[valid summit]

and without this, have the same problem, the sd opens, and shows the log1

[vocal burrow]

I know

[valid summit]

i dont have 7zip, but i'll download

[vocal burrow]

Using Power Shell
1. Press Windows+R to open the Run box
2. Type powershell and click OK.
3. The Windows PowerShell window will open.
4. Run the following command:

Get-FileHash C:\file\path\my_file.exe -Algorithm SHA256

Make sure to replace C:\file\path\my_file.exe with the actual path to the file

[valid summit]

wait what file i have to click?

[vocal burrow]

the model

[vocal burrow]

please read instructions again

[valid summit]

lol im dumb

but im still confused, what file i have to put on that?

[vocal burrow]

7pagNSFWMergedModel_7pagLSCX.safetensors

[valid summit]

okok

like this?

C:\Users\usuario\Documents\Stable Caralhos Duros\stable-diffusion-webui\models\Stable-diffusion\7pagNSFWMergedModel_7pagLSCX.safetensors\my_file.exe -Algorithm SHA256

[vocal burrow]

no

Get-FileHash "C:\Users\usuario\Documents\Stable Caralhos Duros\stable-diffusion-webui\models\Stable-diffusion\7pagNSFWMergedModel_7pagLSCX.safetensors" -Algorithm SHA256

[valid summit]

lol i forgot "Get-filehash"

ok i put on it

[vocal burrow]

and you forgot to remove my_file.exe, and you didn't put double quotes to prevent space from breaking everything

ok so what s the hash ?

[wind sun]

please

[valid summit]

Algorithm       Hash                                                                   Path
---------       ----                                                                   ----
SHA256          DDA4C6CFE3A38B58366C1B76FF9A693E50D71DD37174F916156FB5C778C84F7C       C:\Users\usuario\Documents\Stable Caralhos Duros\stable-diffusion...


PS C:\Users\usuario>
PS C:\Users\usuario>

if got better to read like this

[vocal burrow]

#🤝｜tech-support message or change auto1111's port with --port xxxx

[vocal burrow]

ok so the file is good, does it happen with other models ?

[valid summit]

yes

i've re-installed the sd and stills with the same problem

[vocal burrow]

ok so try to reinstall sdwebui following the pinned guide

And also run pip cache purge beforehands

maybe there was some corrupted cache file and/or the spaces in the path is messing something up

[valid summit]

ok, i'll reinstall sdwebui again with the pinned guide

[wind sun]

dosent work

[wind sun]

where do i do that

[vocal burrow]

it's a command line argument, you put them in webui-user.bat

going to bed then. bye

[wind sun]

dosent work

but alr

Good night

[steady thistle]

AMD is so frustrating.

[lone vessel]

got SD installed locally and running without errors, but the art it generates looks like poop compared to what the bots here generate. is there a specific configuration they are using?

i mean obviously there is, does anyone know what it is and what models etc i need

are there instructions / schematics which would allow me to have an exact copy of the bots here? because that would be a good starting point

[stoic cave]

If you run StableSwarmUI they have official presets you can add. You can also lookup the videos by staff members. They talk about workflows they use internally.

[patent meadow]

RuntimeError: don't know how to restore data location of torch.storage.UntypedStorage (tagged with privateuseone:0)

[lone vessel]

stupid question why does the sever not package the version the bots are using and make it available for download

[peak bay]

hello I am trying to run the stable diffusion webui with my gpu but when I try running it, I notice that my gpu usage is pretty low and it also take a very long time to give the output. Does anyone know what could be the problem?

[lone vessel]

the installation was such a pain in the a$$, i understand the reasons, like its constantly being worked on etc, but i still continute to question if it truly needs to be this hard

[nocturne topaz]

Are you using an amd cpu or a nvidia one?

[lone vessel]

stableswarm does look good im checking it out

[peak bay]

nvidia

[nocturne topaz]

when you look at the gpu usage, click where it says 3d and select cuda.. does it now show heavy usage?

[patent meadow]

i googled this error nothing useful came out

[peak bay]

im using the nvidia-smi command to check gpu memory usage

[nocturne topaz]

windows or linux?

[peak bay]

linux

and its on a vm

[nocturne topaz]

wsl or virtualbox?

[peak bay]

its an instance from gcp

[nocturne topaz]

google cloud.. ok

which gpu is it?

just out of curiosity

[peak bay]

A100 with 40gb of vram

[nocturne topaz]

nice 🙂

[peak bay]

and it was only using 6gb lol

[nocturne topaz]

what size image were you making?

and what was the batch count/batch size?

im assuming this is a1111?

if you were only making 1x 1024x1024 image, then yeah 6gb is about all it'll need

[peak bay]

i was using the wav2lip extension and for that I was using a 4 minute audio/video sample which was estimated to take around 5 hours... which seems slightly unreasonable for A100 lol

[nocturne topaz]

Is there some sort of batch option for that extension (I've not used that one sorry)

[peak bay]

no i dont think so

[nocturne topaz]

4 mins at how many fps?

[peak bay]

~20 fps

[nocturne topaz]

at 24 fps thats 5760 images.. each at say 20 steps is 115200 steps.. what resolution was the video?

[lone vessel]

question about stableswarm: is this something self-contained? or so i install it over top of SD? Because the installation instructions make no mention of how to install it in relation to a pre-existing SD setup

[nocturne topaz]

even at 10it/s thats 192 minutes or over 3 hours, and 10it/s for a decent sized video (1280x720) sounds about right

[peak bay]

360p

[nocturne topaz]

looks self contained to me

[stoic cave]

You can add a pre existing as a backend or setup new

[lone vessel]

ah so is this the easy install SD for dummies i thought didnt exist? 🤓

[lone vessel]

is there a progress bar im not seeing?

i dont see anything indicating how close to being done it is

"Error

No images were generated (all refused, or failed - check server logs for details)."

refused? is this censored?

does stableswarmUI have censorship?

why does it keep generating the same image? ive got seed set to -1

[inland latch]

yo, if someone wants to help me get set up I'll send you some ETH. I got it halfway working, but keep getting errors when I run the main python script. Im trying to do stable video generation from text. I have development experience but having a little trouble

[lone vessel]

i switched on "Video" but so far i dont see it animating. does it just take time?

[random hare]

pardon me, complete newbie and not good at tech at all, but what does can i do to solve this issue just not being able to install / use properly from the start

if this inquiry is too simple/ basic then maybe a DM reply can suffice

[nocturne topaz]

What gpu have you got?

[random hare]

hope this works :>

[nocturne topaz]

Whats the one on the right say for you?

[random hare]

oh yea oops

[nocturne topaz]

is that the amd fork of a1111, or just the standard install (or is it forge)?

[random hare]

is this what you are looking for

[nocturne topaz]

not quite.. how did you install a1111 is the question?

ie. which site did you download it from?

[random hare]

oh if ur talking bout webAI auto 1111, its from a youtube tutorial had good ratio and everything

the main SD site also has a similar tutorial, if its not that then i have no idea honestly sry

[nocturne topaz]

ok.. IMO I think your best bet is probably to create a new folder and install the version from here https://github.com/lllyasviel/stable-diffusion-webui-forge and in the webui-user.bat file set commandline_args=--direct-ml

[random hare]

ok gimme a sec pls, should i delete the old one

[random hare]

is doneso

should i be installing forge too, i guess i will follow all the instructions?

[versed bough]

trying to use the tagger but i keep getting an error it says that wdtagger is no longer maintained

[cosmic spindle]

Hi, I've got an issue with the Extras tab on Automatic1111s webui. It saves the upscaled images as 00001.png. Is there any way to save it with some other filename?

[random hare]

after extracting this package to the same folder as webui-user, and followed these steps, still no response
pls, any ideas?

[nocturne topaz]

add that to the webui-user file right after --direct-ml and try again

[ornate elk]

SD forge is bugged and won't work for amd

[ornate elk]

Pls follow my install guide for AMD.
Its In the pinned messages of this channel

[nocturne topaz]

It was designed for use on AMD - illyasviel is AMD user?

[ornate elk]

I tried it two days ago, with my AMD GPU and it didn't worked

Github users said the same. Maybe he fixed it, would need to test again.

[nocturne topaz]

I've misunderstood..

that isn't illyasviels acct

[ornate elk]

Check the settings under Path/saving

[cosmic spindle]

I did fipple with it, but neither worked 🤔 would you mind sharing the 'correct' setup?

[ornate elk]

Not at the PC rn sry

[cosmic spindle]

Alright

[vague compass]

already added. but i'm still facing 504 in the api if there are multiple users. i have nvidia a10 gpu with 16 cores and 64 gb ram.

[random hare]

installing for AMD, based on the install guide, may i know how to bypass this

[obtuse crag]

🙂

[ornate elk]

The guide says you need to make a new Folder first. You didn't, so thats why you get a folder permission error.
You can right click the folder and under settings security you can give your user and system all permissions.
Or install again. (In a folder you created before)

[obtuse crag]

yall getting on nvidia?

i missed something?

[fossil stone]

I have the same issue.

[lapis merlin]

How do I fix this?

 UserWarning: The operator 'aten::frac.out' is not currently supported on the DML backend and will fall back to run on the CPU. This may have performance implications. (Triggered internally at D:\a\_work\1\s\pytorch-directml-plugin\torch_directml\csrc\dml\dml_cpu_fallback.cpp:17.)
  w = t.frac()

[dull hull]

Hey, haven't used Stable Diff for like 5 days, today after the usual update, generating images is waaay slower than before. It usually took like 20 seconds at max to create 4 images (txt2img), now it is like 1-2 minutes... using 3080 GPU with xformers, everything is up to date, no errors or anything, and doesnt matter which model I use, it is the same.

Is it just me? What happened with the latest update?

[vocal burrow]

v 1.31, that s definitely not "5 days late", that s more like 6 months late

we re on 1.7 by now

[dull hull]

Hmm, that's weird.
Commands are set in webui to auto update A111.. gonna take a look at it then first, thank you

[dull hull]

Aight so the update 1.7 didn't changed it, but removing the --no-half from webui did the trick. (don't remember why that was there atm)
img2img still very fast as usual, txt2img still feels a little bit slower, but not that terrible as before.

Thx for the help tho 🙂

[ornate elk]

make sure you use --medvram-sdxl --no-half-vae if you gonna use SDXL models

[dull hull]

Nah I don't use SDXL, not my cup of tea.
Any tips for further speed-ups tho?

My config is:
Motherboard: ASUS ROG Strix X570-E Gaming
CPU: AMD Ryzen 5 5600X
GPU: NVIDIA GeForce EVGA RTX 3080 FTW3
Memory: HyperX 32GB Fury RGB DDR4 3600MHz CL18

[ornate elk]

you can remove the "call" before git pull

otherwise it wont update

[dull hull]

Oh aight, done. Thx 🙂

[ornate elk]

Make sure if you gonna use Hires Fix that you use 10-15 Hires steps, and not 0

[dull hull]

I usually use img2img, very rarely Hires fix.
These are my "usual" settings at start before I send it to img2img. I've seen others recommended 40-50 steps for 3M SDE Karras due to the details.

Maybe it is a bit overkill 😄

Btw, is Refiner worth using?

[ornate elk]

you said its slow, you should test with batch size 1 and batch count 1 to really see a speedup (or the actual speed) , or using an faster sampler which requires less steps

[ornate elk]

nope, not worth it ^^

[dull hull]

Now done a few testing, with 1 batch size and count. It's still slower for some reason.
One picture took more than half a minute. I've only reduced sampling steps to 40 from 50, and adetailer disabled.

[ornate elk]

can you try an euler a, 30 steps image on 512x512 with a model that is 2gb in size?

[dull hull]

Used Absolutereality (Used Realcartoonrealisticv12 before) now with the settings u mentioned, 27 secs.

But I think I've found the "issue".

When I've removed the given Lora prompt, it took literal seconds to make the picture.

Which is weird, never encountered this thing before...
Gonna check how other Loras will do

[ornate elk]

ah yea, loras and prompts with more tokens than 225 can take (much) longer to generate

[dull hull]

I've done this with 55 pos and 113 negative prompts. Which is still much lower what I usually use

28 seconds with a different Lora with my usual settings (except batch which is 1/1) + ADetailer
Then I guess it was the Lora itself... good to know it can have such great impact 😄

Thanks for the help tho

[ornate elk]

no problem 🙂

do you upscale in img2img later or why you dont use hires fix?

[dull hull]

Well, if I don't like the given created picture or that's not I wanted to do with the prompts, then using hires fix for it is a waste of time.
Plus img2img much faster, and I can play around with photoshop to fix the smaller details faster.

However, in some cases Hires Fix was much better

[ornate elk]

ah okay, yea the time it adds is pretty much, not that good for testing stuff

auto1111 btw has a extension for photoshop, maybe worth a look

[fossil stone]

I'm trying to generate img2img with api but it gives 504 Gateway Time-out error anyone have an idea?

[vocal burrow]

fwiw everything seems to be good with the API at the moment https://dreamstudio.com/api/status/

[fossil stone]

I'm using Automatic1111

[vocal burrow]

what exactly are you trying to do ? And what's the full log ?

[fossil stone]

I want to generate an img2img image by Automatic1111 API, I'm hosting Automatic1111 in AWS sagemaker notebook and I'm sending request to postman but it throws "504 Gateway Time-out"

[karmic crown]

what mac do you have? what COMMANDLINE_ARGS do you have in webui-user.sh?

[karmic crown]

do not do that. up to 36GB of RAM it is 2/3 for VRAM by default, and 3/4 above 36GB

[quiet swift]

I have a mac book Pro from 2021, Apple M1 Max, 32 Giga of Ram. I don't know where I can find the answer to your second question 😦

[karmic crown]

find a folder where is your webui installed and open file named webui-user.sh in editor

[quiet swift]

In the folder webui-user.sh, there is a long text, in which i found that : # Commandline arguments for webui.py, for example: export COMMANDLINE_ARGS="--medvram --opt-split-attention"
#export COMMANDLINE_ARGS=""

python3 executable

#python_cmd="python3"

git executable

#export GIT="git"

[karmic crown]

you need specific params to be able to use a1111 properly on your mac

[quiet swift]

OK. Idid use a1111 without any troubles during many month. Since 15 days or more, it does not work anymore

[karmic crown]

when you run webui.sh, you have something like this in terminal

################################################################
Launching launch.py...
################################################################
Python 3.10.13 (main, Aug 24 2023, 12:59:26) [Clang 15.0.0 (clang-1500.1.0.2.5)]
Version: v1.7.0
Commit hash: cf2772fab0af5573da775e7437e6acdca424f26e

show me what you are getting

[karmic crown]

it can work with default params, but it is not optimal
you probably used some new model, bigger resolution or something like that, so it "stoped" working

also show me what you have on the bottom of user interface in browser, something like this

version: v1.7.0  •  python: 3.10.13  •  torch: 2.1.2  •  xformers: N/A  •  gradio: 3.41.2  •  checkpoint: 52694bdbf4

[quiet swift]

here is what is written when I run webui : ################################################################
Launching launch.py...
################################################################
Python 3.10.12 (main, Jun 20 2023, 19:43:52) [Clang 14.0.3 (clang-1403.0.22.14.1)]
Version: v1.4.1
Commit hash: f865d3e11647dfd6c7b2cdf90dde24680e58acd8
Installing requirements
ControlNet init warning: Unable to install insightface automatically. Please try run pip install insightface manually.

[karmic crown]

ControlNet init warning: Unable to install insightface automatically. Please try run pip install insightface manually.

you have some errors here

Version: v1.4.1

this is your problem

you have ancient version of a1111

do you want fast and easy solution?

[quiet swift]

here is what is written on the bottom of SD : API  •  Github  •  Gradio  •  Startup profile  •  Reload UI

version: v1.4.1  •  python: 3.10.12  •  torch: 2.0.1  •  xformers: N/A  •  gradio: 3.32.0  •  checkpoint: bff4610d23

[quiet swift]

Yes, I want fast and easy solution 🙂 🙂

[karmic crown]

where is your a1111 installed? is it in stable-diffusion-webui in your home folder or somewhere else?

[quiet swift]

I don't know where a1111 is installed. I think that everything is installed inside this folder :

[karmic crown]

do you know how to do that?

[quiet swift]

no, 😦

[karmic crown]

right click (or crtl+click) depending on your configuratuon, then go to services and chose new terminal at folder

like in the picture i sent

[quiet swift]

ok, I did that. Nouveau terminal au dossier, which must be new terminal at Folder

[karmic crown]

type pwd and send me the output

pwd (and then press enter)

❯ pwd
/Users/viking/stable-diffusion-webui

it will show your folder (that is mine)

[quiet swift]

arnaudbigeard@MacBook-Pro-2 IA % pwd
/Users/arnaudbigeard/Desktop/IA
arnaudbigeard@MacBook-Pro-2 IA %

[karmic crown]

that is not the folder you should me

[quiet swift]

oh, sorry

[karmic crown]

this one

[quiet swift]

Ok I understood : arnaudbigeard@MacBook-Pro-2 stable-diffusion-webui % pwd
/Users/arnaudbigeard/stable-diffusion-webui
arnaudbigeard@MacBook-Pro-2 stable-diffusion-webui %

[karmic crown]

ok

that is even better

❯ git status
On branch master
Your branch is up to date with 'origin/master'.

type git status and press enter

and show me the result

like that

[quiet swift]

Sur la branche master
Votre branche est à jour avec 'origin/master'.

Fichiers non suivis:
(utilisez "git add <fichier>..." pour inclure dans ce qui sera validé)
.DS_Store
.github/.DS_Store
deforum_settings.txt
extensions-builtin/.DS_Store
modules/.DS_Store

aucune modification ajoutée à la validation mais des fichiers non suivis sont présents (utilisez "git add" pour les suivre)
arnaudbigeard@MacBook-Pro-2 stable-diffusion-webui %

[quiet swift]

Sorry, everything is in french !! 🙂

[karmic crown]

give me a second, something is wrong my translation app

i am not logged in

that looks fine

that one looks up to date

do not run the other one

run that one

type

./webui.sh

and press enter

you should have something like this in terminal

################################################################
Launching launch.py...
################################################################
Python 3.10.13 (main, Aug 24 2023, 12:59:26) [Clang 15.0.0 (clang-1500.1.0.2.5)]
Version: v1.7.0
Commit hash: cf2772fab0af5573da775e7437e6acdca424f26e

with 3.10.12

[quiet swift]

here is what I have : ################################################################
Install script for stable-diffusion + Web UI
Tested on Debian 11 (Bullseye)
################################################################

################################################################
Running on arnaudbigeard user
################################################################

################################################################
Repo already cloned, using it as install directory
################################################################

################################################################
Create and activate python venv
################################################################

################################################################
Launching launch.py...
################################################################
Python 3.10.12 (main, Jun 20 2023, 19:43:52) [Clang 14.0.3 (clang-1403.0.22.14.1)]
Version: v1.4.1
Commit hash: f865d3e11647dfd6c7b2cdf90dde24680e58acd8
Installing requirements
ControlNet init warning: Unable to install insightface automatically. Please try run pip install insightface manually.

[karmic crown]

?!

how did you run it?

have you type ./webui.sh in the termina?

[quiet swift]

I've got many other text under what I did copy paste for you

[quiet swift]

Yes I did : aucune modification ajoutée à la validation mais des fichiers non suivis sont présents (utilisez "git add" pour les suivre)
arnaudbigeard@MacBook-Pro-2 stable-diffusion-webui % ./webui.sh

################################################################
Install script for stable-diffusion + Web UI
Tested on Debian 11 (Bullseye)
################################################################

################################################################
Running on arnaudbigeard user
################################################################

################################################################
Repo already cloned, using it as install directory
################################################################

################################################################
Create and activate python venv
################################################################

################################################################
Launching launch.py...
################################################################
Python 3.10.12 (main, Jun 20 2023, 19:43:52) [Clang 14.0.3 (clang-1403.0.22.14.1)]
Version: v1.4.1
Commit hash: f865d3e11647dfd6c7b2cdf90dde24680e58acd8
Installing requirements
ControlNet init warning: Unable to install insightface automatically. Please try run pip install insightface manually.

[karmic crown]

ok

https://github.com/viking1304/a1111-setup/releases/tag/0.0.5

download my script v0.0.5

put it wherever you want, just do not put it in your stable difusion folder

[quiet swift]

ok

[karmic crown]

extract and open that folder in terminal

run it like this

sh a1111-setup.sh -f all

it should fix your installation in your home folder
this one
/Users/arnaudbigeard/stable-diffusion-webui

[quiet swift]

I have downloaded your file :

[karmic crown]

open this in terminal

and run it like i wrote

[quiet swift]

but I do not know how I can open that in terminal, sorry

[karmic crown]

"ok, I did that. Nouveau terminal au dossier, which must be new terminal at Folder" <- like you did here

[quiet swift]

OK

[karmic crown]

be sure to close other instances of webui first

best close the terminal

[quiet swift]

I did close terminal and Stable Diffusion. But I do not fucking succed to do the "new terminal at folder" 😦

[karmic crown]

do you use right click or ctrl+click?

one of those will give you secondary menu

go to services

and there is the option (in french) Nouveau terminal au dossier

mine is in enlish

[quiet swift]

for a strange reason, it tells me only teamwiever 😦

[karmic crown]

rapid is quick 🙂

hmm

that is quick actions

is it your downloads?

when you select this

what you have at the bottom

i know

do this

[quiet swift]

Yes, when I select this, and do Nouveau terminal au dossier, that did open a new window in terminal

[karmic crown]

ok

now run it like
sh a1111-setup.sh -f all

[quiet swift]

Terminal is writting many things.....

[karmic crown]

it should start like this

it will first check if you have all required things (python, git..)

[quiet swift]

it begins like this :Checking for required packages...
Installing cmake...
cmake is already installed
Installing protobuf...
protobuf is already installed
Installing rust...
rust is already installed
Installing python@3.10...
python@3.10 is already installed
Installing git...
git is already installed
Installing wget...
wget is already installed

Updating A1111 installation in /Users/arnaudbigeard/stable-diffusion-webui

HEAD est maintenant à f865d3e1 add changelog for 1.4.1
remote: Enumerating objects: 7826, done.
remote: Counting objects: 100% (1987/1987), done.
remote: Compressing objects: 100% (68/68), done.
remote: Total 7826 (delta 1937), reused 1964 (delta 1919), pack-reused 5839
Réception d'objets: 100% (7826/7826), 2.92 Mio | 24.13 Mio/s, fait.
Résolution des deltas: 100% (5491/5491), complété avec 191 objets locaux.
Depuis https://github.com/AUTOMATIC1111/stable-diffusion-webui
f865d3e1..cf2772fa master -> origin/master

• [nouvelle branche] bump-Pillow-blendmodes-dependency -> origin/bump-Pillow-blendmodes-dependency
• [nouvelle branche] compact-Soft-Inpainting-infotext -> origin/compact-Soft-Inpainting-infotext
• [nouvelle branche] consistency-dec -> origin/consistency-dec
  11f33973..b7f45e67 dev -> origin/dev

[karmic crown]

looks fine so faar

[quiet swift]

and it ends like this :
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
Warning: caught exception 'Torch not compiled with CUDA enabled', memory monitor disabled

[quiet swift]

ouahh, Stable Diffusion did open like that 🙂

[karmic crown]

that is the momenet when my script ends

and a1111 starts

let me know when it opens

and show me what you have at the bottom

[quiet swift]

Stable Diffusion is open and at the bottom : version: v1.7.0  •  python: 3.10.12  •  torch: 2.2.0  •  xformers: N/A  •  gradio: 3.41.2  •  checkpoint: bff4610d23

[karmic crown]

that is good

next time just find this folder in your home folder

[quiet swift]

Wonderfull. Thank you so much. I've got a last question. When I need to open SD, I open Terminal and then I write : "cd ~/stable-diffusion-webui;./webui.sh" Is it ok ?

[karmic crown]

and click this

this is faster and esier 🙂

close it and try

[quiet swift]

Ok

[quiet swift]

I also close Terminal ?

[karmic crown]

yes

that will close stable diffusion

[quiet swift]

I have closed terminal, closed SD, and then I did double click on the folder webui.sh, but it just open in Text Edit. I must do something else ?

[karmic crown]

open with terminal

[quiet swift]

ok

[karmic crown]

it should work

if not

open terminal

and type

cd
cd stable-diffusion-webui
./webui.sh

[quiet swift]

ok

[karmic crown]

cd
cd stable-diffusion-webui

is the same as

cd ~/stable-diffusion-webui

just easier if you have no idea where ~ is on your keyboard 🙂

[quiet swift]

It works. Yes 🙂 thank you !!

[quiet swift]

yes, it's strange that I can not find this quick access to "new terminal at folder" 😦

[karmic crown]

cd without any params will get you in your home folder

that is something i use a lot

[quiet swift]

I must leave home, but thank you again for your help !! I use a lot Deforum to create strange and spooky slow videos 🙂 And then I mix Deforum with RunWayGen2 and other AI to create slow movements.

[karmic crown]

i am glad i could help

when you want to update, just go to terminal, go that folder and type
git pull

to update

sorry

i am to tired, long day at work

[quiet swift]

git pull ok 🙂

[karmic crown]

bye guys see you, i will have hell at work tomorrow again i need some rest

[quiet swift]

good night

[nova ermine]

I'm AI generating this character, but mouth is horrible. I thing SD doesn't understand that's the mouth. How to fix?

I'm using this for img2img

And this in ControlNet lineart realistic

[nocturne topaz]

id get out photoshop, crudely paint in the mouth you want then use img2img to fix it

[scenic saddle]

Hey when I open the webui-user.bat file and try to run it to gain access to stable diffusion interface it gives me this error, Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check, any ideas on how to fix this

[ornate elk]

hey, whats your GPU?

[scenic saddle]

intel (R) iris (R) xe graphics, aka intel

[flat olive]

im here

[ornate elk]

then you need an other webui

the default one is for nvidia only

[flat olive]

how i can tell you which version im ussing?

[ornate elk]

whats your webui folder named?

[scenic saddle]

ahh ok thank you very much

[flat olive]

D:\Stable DIfussion\stable-diffusion-webui

i tried de directml

but nothing

[flat olive]

it doesn't work for me

[ornate elk]

you need automatic1111 with Openvino for Intel Iris XE support

[ornate elk]

it should work if you follow my AMD install tutorial, its in the pinned message of this channel

[flat olive]

oky

[ornate elk]

make sure to do the important step mentioned there

[flat olive]

ty

i let you know if it dosen't work?

[ornate elk]

sure

you can delete the other webui

it was the one for nvidia

[flat olive]

heeey

i try to generate but that thing show

i fix it

i used --disable-model-loading-ram-optimization on the args

it works

ty @ornate elk

[ornate elk]

Perfect, no problem 🙂

[ornate elk]

Do you have 8gb RAM ?

[flat olive]

yep

[ornate elk]

Ah that explains the error

[flat olive]

it is now working with my gpu and its 10x more faster

do you have a youtube video with that installation?

cuz it could be really nice for the people like me, my first option was github and yt

[ornate elk]

Ah, but nope didn't saw any useful guide like mine on yt.
But I will add it to my github info wiki

Its pretty difficult for amd users as there are many outdated guides :/
Or guides that dont cover everything.

[flat olive]

yep

i see some post of 6months

[ornate elk]

Yea, and even the "official" guide from AMD is outdated and promotes the not as good configuration

Stuff changes fast in Ai time. Also most AMD tutorials broke after the webui update changed the launch arguments needed to start.

[wind sun]

please someone help me with that

yall keep ignoring me

[nocturne topaz]

you tried zluda yet?

[old bane]

trying to help a friend of mine out getting Stable diffusion and we got very close but its giving out red errors like crazy after trying to launch the webui. we are using stable diffusion XL with automatic 1111
this is the screen shots they sent me but im out of depth with this stuff. please help

[nocturne topaz]

They can't download anything in python.. are they on a vpn or have some sort of wacky networking solution?

[old bane]

ill ask

[old bane]

they said they didnt have there VPN on at the moment but dose have express vpn on there pc.

[old bane]

still looking for help. please. can anyone give me technical support here?

[upbeat crane]

kinda hard to make out whats in your screenshot

i see a good starting point would be to update your pip

[old bane]

to the most resent version?

[upbeat crane]

yeah

i recommend following sebastian kamph's tutorial on installing

https://youtu.be/kqXpAKVQDNU?si=3wEEEHKx1gwlyfjf

might even start over and just take your time, making sure not to miss any steps

sebastian does a pretty good thorough job of explaining the install

this video is for a local install btw

[old bane]

we are atempting to do a local install

[upbeat crane]

once you have that installed you will just need to pull the SDXL checkpoint and refiner found on huggingface

then you can start pulling models from civitAI and such

yeah i would basically start over then (or just start a new install folder)

follow along with sebastian's video, and i think you will get it done

[old bane]

ya... i was trying to help them from a install tutorial i did from like a year ago or something. before XL and me trying to do the instructions step by step have not worked even slightly!

[upbeat crane]

it also can be important what version of python you have installed, i know ive had issues with certain things unless i had 3.10.6 installed

just want to pay close attention there, and when you install python i think its good to install it to the root of c: or somewhere easy like that

[old bane]

hmm?

[upbeat crane]

have you installed python?

its all good if not that video will walk you though it

[old bane]

its a friend of mine trying to install not me. i have a working version lol

[upbeat crane]

make sure you add python to system path, if you arent sure of that then just google it so that python is available system-wide

cool, i would just tell him to start from scratch using that instructional video then

[old bane]

ok there currently downloading. they have a slow internet but i will AT you for an update. also thanks for helping out so far

[upbeat crane]

im no pro but i will help out where i can

kinda looks like he skipped out on some preliminary steps before trying install but the video covers them

im sure he will get it

[rugged latch]

ngl, to an outsider, ik from experience, it looks like stable is so flippen hard to instal, like all these steps, things to make sure are there, all that nonsense

took me nearly a month to work up the courage to try

and when i did, it was so easy, u get the hang of it and soon u able to do it mote than twice, i had to redo it once and now i have 2 stable versions installed, one for cpu and one for gpu xD

[upbeat crane]

ive bricked so many installs now that i couldn't find solutions for that i finally learned to create backups as i go using 7-zip or winrar, etc.

that way i don't have to start from scratch when i do something silly that borks my install lol

[rugged latch]

yesh

smort

[upbeat crane]

i've been using SD for about a year and feel like i have a decent grip on quite a few things, but i have always kinda skipped over inpainting and outpainting, however i have attempted on several occasions and get terrible results (even while following along with some of the most basic tutorials). i have never seen any tutorial mention anything about utilizing specific 'inpainting models' and am just a bit confused about this aspect. i'll pull a model from CivitAI and be trying to do something basic like create a mask and inpaint and it just NOT work at all.

[rugged latch]

i just copied and pasted my first instal to an ssd, and after i updated my orgianl copy to like, ,idk, wut ever the recentt version is, and have it installd for my gpu, i now have a legacy verison of stable for cpu and newiest for gmpu, oddly enough with same prompts and models and settings, the new version of stable for gpu makes diff pics nu matter wut, sure i can edit the prompts to make nearly the same result but still odd and worth noting, gpu seems to follow prompts less..idk, less acuate ig i can say, usually requireing me to be extremly specifc on stuff

whereas tje legacy cpu veriosn pretty much nails it each time

[upbeat crane]

i'll also see sometimes a specific model on CivitAI and then you will also see there is that same model but Inpainting after the name. I am wondering, do i need that model in ADDITION TO? or are those usually the full model that just happen to be better trained on Inpainting. think you see where my issues are now. also, i see quite a few ppl discussing the Inpaint Anything extention, wondering what route people take when they have issues with inpainting doing what they are telling it to do, do they try and pull a specific inpainting model for whatever checkpoing they used, just use an extension like i mentioned, etc.

yeah that is interesting hmm

[rugged latch]

im assuming, its just how the AMD 5600G cpu is designed at the waffer level

as to why the cpu portion of the dye handles prompts diffirently than the graphics portion

[upbeat crane]

appreciate any and all feedback on the inpainting question anyone has to offer. i know it's pretty basic stuff but i am seriously stumped as to why i have never had much luck with inpainting, as i have obviously missed the mark here somehow, much appreciation!

[rugged latch]

i say just tinker wit it a bit

xD

[upbeat crane]

done a LOT of tinkering and watched many tutorials and nothing ever matches up to what they are showing.

its very odd, like i can create animations using deforum, so im good with this stuff generally, i think im just missing something major here and i cant figure it out

[nocturne topaz]

if you inpaint, then yes you'll want the inpaint version of the model, otherwise if just doing txt2img and img2img then you won't need it

[upbeat crane]

gotcha

so, say there is CHECKPOINTXYZ on civitAI, then you also see CHECKPOINTXYZ Inpainting? would the the Inpainting one be the FULL model but just be better trained on inpainting

or would i somehow need to be utilizing BOTH in some way?

[rugged latch]

i think, based wut chief said

just use the inpainting

[upbeat crane]

yeah, so from now on i will always see if an INPAINTING checkpoint is available and then go for THAT one, and if there isnt a specific Inpainting checkpoint then just go with whatever base is available. that sounds reasonable and i can work with that

[rugged latch]

like basically, if u arew u sing txt2img or img2img, u would want to use theh regualar verison, and if inpaint, u wanna use the inpaint copyt

also, lora values can go down as far as negative integers? ovO

[upbeat crane]

yeah except maybe you wouldnt know you wanted to inpaint something until days/weeks later, you may have just used the base model then tried to inpaint on an image and it doesnt work

i will work with this knowledge for now and keep digging and playing and ill come back if i have further issues

appreciate the feedback gang

[rugged latch]

ofc :3

[upbeat crane]

yeah they can on the lora question

[rugged latch]

whaa

[upbeat crane]

you can place lesser weights on a lora using negative integers

[rugged latch]

how does that make senseee

[upbeat crane]

i use some loras that are just TOO strong, and i like to back off on them a bit and get way better results, especially when combinging with multiple loras

[rugged latch]

i thought that with a lora at 0.0, it wouldnt appear at all

[upbeat crane]

using a .8 for example would back off the full value of the lora by my experience

oh maybe i got confused, i dont use - per se

but rather numbers LESS than one. .8, .7, etc

[rugged latch]

in my experience, loras at 1.0 was full blast, 0..8 to 0.1 was always, lesser and lesser, and by the time u get to 0.1 or 0, u didnt see it at all, so it cemented itself in my head as, 1.0 is 100% and any lesser would be like, and % less than 100

[upbeat crane]

that is also my understanding, sorry if i created confusion

[rugged latch]

<v> Ya and when cheif said they can go as far up as 5 and as far down as -5, ii got really confused

[upbeat crane]

oh i see

ive also noticed that some loras have a certain value, even as low as say....2.4 that just blow out the image with noise

[rugged latch]

i just dunt see how, code wise, a negaitive numeber can mean anything to amount of something less than 0

[upbeat crane]

so they are all different i guess

[rugged latch]

as 0 is well, bedrock level

liie, mathematically speaking

just dunt make sense

[upbeat crane]

yeah i have no understanding of a negative integer for a lora weight, so dunno

[rugged latch]

xD ikr

same

brain goes brr

[upbeat crane]

be like saying, i am adding this lora prompt here, but whatever you do do NOTHING like what this lora does lol

[rugged latch]

like, if a lora can go to -5, thats implying that the medium is 0, which cant be, or thatd mean that the values are in definition, infinite

and that cant be

[upbeat crane]

so let's test it then

[rugged latch]

xD

i thihk i tried a negative by accident once and i swear it throwed a runtime error

and said sum around the lines of, invalid value or sum

[upbeat crane]

it took the - weight but ill have to play around more and use the same seed and compare/contrast

interesting question tho

[lone vessel]

can i get a real answer to this?

[tender nova]

Anyone here?

Bot1 etc not listed cant make ai gen photos

[old bane]

@upbeat crane they sent me this screen shot.

[fossil dew]

I downloaded these both via inastal from URL and I can't seem to figure out how to get them to work?

[lunar linden]

anybody know why auto1111 /sdapi/v1/img2img returns this error "'str' object has no attribute 'convert'" when a mask is passed through? request is formatted correctly, it works on my local machine but not on a remote server

[sonic fable]

Can someone tell me the differences between these models? Or refer me to a documentation pls
dpmpp_2m_sde
dpmpp_2m_sde_gpu
dpmpp_3m_sde
dpmpp_3m_sde_gpu

[fossil dew]

ok 😐 Now I installed the stable-diffusion-webui-wildcards

but the file is somewhere - I found it, but

Do I copy all of the files into the stable diffusion XL > webui folder?

cuz its not showing up when I restart the stable diffusion UI

and putting it in the webui folder makes the most sensee- but im not big brain with this stuff (i useually ask my partner)

[upbeat crane]

https://stable-diffusion-art.com/samplers/ @sonic fable

[fossil dew]

EXTENTIONS THEYRE IN MY EXTENTIONS FOLDER ok?! but why aren't they showing up when I restart the UI?

[upbeat crane]

close and reopen your python launcher window too maybe?

[fossil dew]

i tried that- but I will try it one more time

[upbeat crane]

make sure that all instances of the launcher and the ui are closed

[fossil dew]

nuffin

[upbeat crane]

never used this particular extension, did you install from URL from the extensions tab?

[fossil dew]

yes

Then went to available and clicked install

[upbeat crane]

maybe try to go the git clone route. err since it's already installed try a git pull from the install folder

[fossil dew]

i have NO idea what you just said 😄

[upbeat crane]

https://github.com/AUTOMATIC1111/stable-diffusion-webui-wildcards

look at the instructions on the github page for this extension you are working with

[fossil dew]

OK you mean do it manually?

[upbeat crane]

yeah you could attempt the manual install method using git

but they didnt leave much in the way of install instructions

[fossil dew]

should I delete the ones in the extentions folder first or

i mean- the files are there- i just dont undestand why they're not showin' up

[upbeat crane]

i probably would uncheck the extension at first then go to that folder with your cmd prompt and attempt the git clone

close out of your ui and launcher first then try that maybe

good luck, not really sure what could be jamming you up.

[fossil dew]

ill try that

I appreciate it

[upbeat crane]

yw

do you have sd-dynamic-prompts installed?

@fossil dew

try this, if not https://github.com/adieyal/sd-dynamic-prompts

[scenic saddle]

Hey y'all dose any one know how to fix this problem, python3: can't open file '/content/gdrive//sd/stable-diffusion-w/webui.py': [Errno 2] No such file or directory

[upbeat crane]

sorry i run locally but is it supposed to have the double // after your gDrive location?

https://github.com/TheLastBen/fast-stable-diffusion/issues/2679\