🤝｜tech-support

[karmic crown]

they should be here

[little echo]

yea i did that

[karmic crown]

https://huggingface.co/stabilityai/stable-zero123/tree/main

try those

[karmic crown]

https://www.youtube.com/watch?v=jaRr5W80N8E

this might help

i have never used that

[little echo]

i guess that's the CLI version, might try some time

dissappointed they closed the channel for it here because there's hardly any discussion places for it

[opal bluff]

Thats okay. i have a dual xeon 256gb of ram and 3 Tesla P40s so i wanted to try to get the most ot of all of them.

[proper pike]

What can I use to replace moondream interrogator and show text so I can use my own text as a description?

[misty schooner]

Can I run stable diffusion on my GTX 950 2GB and ryzen 5 3600?

[noble belfry]

you might be able to run a stable diffusion 1.5 module on 2GB

[boreal bison]

guys what's this error in a1111?

I get this after updating

anyone?

[ornate elk]

Hey

When do you get that error

[boreal bison]

hi, when I try to generate something

or do image to image

[ornate elk]

With sdxl?

What's in your webui-user.bat?

[boreal bison]

Which one is that?

[ornate elk]

The one where the filetype is windows batch file

[boreal bison]

@echo off

set PYTHON="C:\Users\logic\AppData\Local\Programs\Python\Python310\python.exe"
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--xformers --no-half-vae --device-id 0
git pull
call webui.bat

[vocal burrow]

screenshot your whole webpage

[vocal burrow]

webpage

so we can see your settings

[boreal bison]

other models seem to work, except juggernautXL

[vocal burrow]

what's in that controlnet tab ?

[boreal bison]

nothing everything is disabled

[onyx solar]

What size is your model file?

[vocal burrow]

ok, try updating your extensions then

[boreal bison]

6.9GB

I did 30 mins ago

doing it again

[vocal burrow]

you can use txt2img fine with this model ?

[boreal bison]

nope

nothing works with this model

[onyx solar]

Have you tried re-downloading?

[boreal bison]

nope

[onyx solar]

Seems like a bad model file to me.

[boreal bison]

I should get the latest version probably too

[onyx solar]

Yeah, 9 is good.

[boreal bison]

It was working fine 😦

any special tricks for v9? or it's the same

[onyx solar]

Same same.

[boreal bison]

downloading

[onyx solar]

Do other XL models work for you?

[boreal bison]

yes

[vocal burrow]

hash does match with what's on civitai so the model is fine

[boreal bison]

which one should I get? lightning?

[onyx solar]

Have you completely shut down and restarted A1111?

[boreal bison]

@onyx solar yes

the problems started occuring after I lost my C: drive

reinstalled windows etc

[vocal burrow]

lighting is faster but quality takes a hit

[onyx solar]

Regular model would be better than lightning

[boreal bison]

I see I'll get both haha

[vocal burrow]

also make sure that id 0 is your 3090 and not your 3080

(weird setup btw)

[boreal bison]

I was assuming it's weird, but I don't know why haha

my 3090 is in the second slot, but it's id 0 seems like

[vocal burrow]

looks like it to me

[boreal bison]

My 3080 is stuck in the motherboard, can't take it out even to sell

[vocal burrow]

can you delete the venv folder and try again ?

Also please provide a full console log, not just screenshot.

[onyx solar]

Stuck in the motherboard?

[vocal burrow]

copy paste the whole log in a .txt and drop it in there

[boreal bison]

yeah

[boreal bison]

let me try

[onyx solar]

Did you push down the release that is at the back end of the slot?

[boreal bison]

ok

deleting vnev folder now

will you help me bring back my settings later? haha

[vocal burrow]

your settings are safe

[boreal bison]

ralaunching webui

it's doing things

[vocal burrow]

yup let it do its things

you had torch 2.0.1, weird for auto1111 1.8.0

[onyx solar]

What motherboard do you have in your PC?

[boreal bison]

I've been away from stable diffusion for 3 weeks probably, is anything new happening?

@onyx solar I don't remember

[boreal bison]

will this get the correct version now?

[onyx solar]

Modern PCIe slots have a release you have to press down on to get the card out of the slot.

[vocal burrow]

it's getting as you speak

[boreal bison]

I know, it's unpressable, broke or something

😄

Anyone living in Glendale, CA feel free to come by and take it out lol

can I get 4090 if I sell my 3090 and 3080? how are the prices now?

[vocal burrow]

definitely not :p

[boreal bison]

got my 3090 overpriced for 2500 back than

[boreal bison]

not good not good lol

[vocal burrow]

4090 is still "I hope you're born with 3 kidneys" price level.

[boreal bison]

snap

just checked, it's not too far off

I could sell both add a few hundred and get 4090 prolly

[vocal burrow]

I mean sure

[boreal bison]

it's probably not worth it though

[vocal burrow]

if you have "a few hundreds" lying around

[boreal bison]

haha

[onyx solar]

Honestly, I think you need to try to remove that 3080 just for my sanity. If the clip was broken, you'd be able to just pull the card out of the slot. The clip is pushed downward, toward the motherboard to release.

[boreal bison]

come to my office take it out, I tried

I mean sure I can use all my strength, but I'm sure if I do that, the earth will change it's orbit axis

[onyx solar]

I'm halfway across the country...but I assure you, if the clip is there it's probably that you didn't push the clip down enough.

[boreal bison]

I will break everything

I can take out my 3090 easily

[onyx solar]

Are you sure you unscrewed both sets of screws on the mounting bracket where things plug in?

[boreal bison]

I don't remember, I don't know

My friend built it for me

[onyx solar]

I've literally worked on hundreds of performance PCs in my life...those are typically the only things I've run into that would stop you from removing the card.

[vocal burrow]

how is that auto1111 venv rebuild going ?

[boreal bison]

yeah I don't even know what screws are you talking about, but maybe we should go live and try it 😄

[boreal bison]

still going

[onyx solar]

The ones that hold the card in the case at the back plate.

[vocal burrow]

very close to the end

[boreal bison]

I don't wanna touch my PC anymore. I live it be as is, I'm afrain to f it up

it's making noises and stuff

[onyx solar]

Your PC is closer to Lego than a complex machine at this point. It's actually rather difficult to fuck it up.

[boreal bison]

hahaha

[vocal burrow]

soul be like

https://tenor.com/view/king-arthur-sword-in-the-stone-excalibur-sword-magic-gif-17323244

tadaaa

[boreal bison]

lmaoooo

[vocal burrow]

free 3080

[onyx solar]

Exactly

[boreal bison]

@vocal burrow exatcly!

only the worthy can pull it out

btw is sd3 out yet?

[onyx solar]

Not fully...very limited group has access right now.

Just a handful of people.

[vocal burrow]

closed beta for now

[boreal bison]

ok still.. gonna wait

@vocal burrow ok restarting webui

stuck here

[vocal burrow]

let it be, give it a minute

[onyx solar]

Patience, Daniel-san.

[boreal bison]

🥳

error with JN8 again

[vocal burrow]

something is wrong with mmyolo's adetailer stuff but let's ignore it for now

[boreal bison]

even if I disable it give me an error

[vocal burrow]

okay let's not ignore it then

[boreal bison]

trying jn9

now

[vocal burrow]

can you disable the extension altogether in the extension tab ?

[boreal bison]

let me try

JG9 works btw

[vocal burrow]

restart sdwebui afterwards

[onyx solar]

Still seems like the 8 file has an issue, especially if 9 works.

[vocal burrow]

urgh that's just weirder then

[boreal bison]

restarting (reloading UI)

[vocal burrow]

but it's 2:00 am and I need a freaking shower.

[onyx solar]

Just because the hash matches doesn't mean the file is perfect.

lol

[vocal burrow]

You caught me right after my workout with the "disable-nan-check" stuff

that just triggered me enough to sit here and deal with it

[onyx solar]

😄

[boreal bison]

😂

[onyx solar]

No wonder it stinks here.

[boreal bison]

disabled adetailer from extensions

ok screw JN8 I hope that's the only thing that doesn't work

[vocal burrow]

tldr btw: disable-nan-check is just gonna remove some "safety check" from SD's pipeline.... So in case there's a problem with the generation it will ignore them and continue as if nothing happened.... sure you won't get an error.... but also 99% you also won't get an output

[boreal bison]

I'm not gonna disable that

[vocal burrow]

So you'll be sitting here looking at a totally black output, without any error message, with nothing to investigate

Every time someone suggest using --disable-nan-check it adds 1 day of delay to the release of sd3

there I spilled the beans

[boreal bison]

hahaha

JG9 yeyyeyey

[vocal burrow]

tbh that's what I'm afraid of.

[boreal bison]

perfect adetailing the screwed face

[vocal burrow]

if it was only a JN8 issue then .... weird but ok I can live with it....

[onyx solar]

Well, that was fun...lost power due to a lightning strike for a moment.

[boreal bison]

if anything comes up, I will let you know

[vocal burrow]

but I'm afraid we'll see each other again in a couple of days with some similar issues

[boreal bison]

hahaha

before you go

any cool things with SD in the past 3 weeks I was missing?

any new cool tools?

[vocal burrow]

not much, there's some cool papers but nothing big has been released.

the spotlight is on sd3

[boreal bison]

any update to Instant_ID maybe?

[vocal burrow]

it's been released for quite a while by now

[boreal bison]

I was hoping to get Instand ID with higher CFG scale

so I don't get boring pictures

[foggy oriole]

Most of the official manuals for AMD involves --disable-nan-check as a remedy at some point, don't blame the common folk for that 😂

[boreal bison]

AMD cpu?

[foggy oriole]

GPU, you are safe, don't worry

[boreal bison]

🙂

[vocal burrow]

Those AMD install instructions haven't been updated for months if not a year by now.

Check pinned for more updated ones. (there's two AMD guides over there, the zluda being the more complex but fastest option)

The only legit usage I can image for --disable-nan-check if is someone is running a metric ton of generation. Let's say he launched a deforum batch for 1000 frames.

And then they let their computer do the work at night.

Then yeah ok maybe... why not.... you don't want it to stop at the first error.

[dreamy carbon]

its like this

and I still get the issue

someone told me this too

but idk how to find the torch and cuda version

[soft fiber]

if youre using a1 then its at the very bottom

looks like this

[noble belfry]

i think there is a torch 2.1/cuda121 binary, but yeah its a losing game keeping up with xformers. im constantly gushing about forge, but if you've a low end pc the optimizations in forge over a1111 make xformers pointless and i think in general its a bit "old hat" now. the issue is, old guides send you to old install links etc and what was bleeding edge 8 months ago is slow as hell compared to modern things

[ember hull]

sometimes the face detailer just outputs a black square, is there a solution?

[proud yarrow]

Does anyone know of a way (maybe with a plugin) that I can change how the search for node functionality works in comfyui (double right click).

I am trying to create workflows using a wearing a VR Headset, but because the keyboard is separate in VR and makes the mouse move out of the search box bounds it will always disappear before I can use my VR keyboard.

It would be nice if there was a setting to disable it disappearing when the mouse goes out of the bounds.

My workaround for now is just learning where all the nodes are in the node tree, or by using the suggested nodes when you drag on an output/input.

[dreamy carbon]

what app is this looks so good

[balmy shore]

when trying to img2img,
resize: from 768x432 to 960x540
using 1.25 scale
it returns
TypeError: 'NoneType' object is not iterable
Time taken: 8.4 sec.

but it works at 1.5 scale, once that is done i can use 1.25 with no issue

seems that resolution should be fine im not sure what the problem is

File "D:\SD-ForgeUI\system\python\lib\site-packages\torch\utils\_contextlib.py", line 115, in decorate_context return func(*args, **kwargs) File "D:\SD-ForgeUI\webui\ldm_patched\modules\utils.py", line 439, in tiled_scale out[:,:,round(y*upscale_amount):round((y+tile_y)*upscale_amount),round(x*upscale_amount):round((x+tile_x)*upscale_amount)] += ps * mask RuntimeError: The size of tensor a (12) must match the size of tensor b (11) at non-singleton dimension 2 The size of tensor a (12) must match the size of tensor b (11) at non-singleton dimension 2

sounds like some type of resolution rounding error but they are even numbers ><

disabling tiled VAE seems to fix the issue,, SD-forge

[valid gorge]

Sorry a bit off topic, anyone has experience using and installing QWEN-VL locally on Windows? I could use some help.

[vocal burrow]

try to stick with multiples of 8 (540/8 = 67.5) => go with 544

[dreamy carbon]

https://www.youtube.com/watch?v=1Ef_nu7DGwk The video im using sorry if this is advertisement. But the first 2 step, pip update and xformer just doesnt work and it hurts

[vocal burrow]

to ugrade pip (not needed), you just have to copy paste the green line it gives you at startup when warning you about a pip update. not sure why he made it more complicated than it needed to be.

and for xformers I need more informations than just "it doesnt work"

[balmy shore]

ok I figured this had something to do with it, was hoping multiples of 2 would be enough

or like I'll tuse manual resolution instead of the slider hmm

although just disabling the never oom tired vae thing worked

tiled*

[ornate elk]

whats in your webui-user.bat?

and whats the error you get?

[vocal burrow]

they never answered me so I think they figured it out :p

[dreamy carbon]

I tried the --xreforms still give me no module

then I tried the the update pip

and it said I didnt have it

[ornate elk]

can you screenshot the webui-user.bat? and tell your gpu model

[dreamy carbon]

NVIDIA GeForce 2060

RTX

[ornate elk]

you also need to add --medvram --no-half-vae
after that launch the webui-user.bat

[dreamy carbon]

sorry I didnt see I assume people pings. I don't understand the upgrade, it just says I dont have it installed

?

[vocal burrow]

yes

that s not gonna fix your "xformers issue" tho

start your sdwebui and give is the full console log (put it in a .txt and drop said file in here)

[ornate elk]

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

[ornate elk]

looks good

[dreamy carbon]

got it

do I only start the user

not the other?

[ornate elk]

yes always the webui-user.bat

[dreamy carbon]

ok ty

the next step is pip upgrade

what am I suppose to do with that

you explained it, but I didnt quite get it

[ornate elk]

you should install python 3.10.11 64 bit first, then the pip will be updated too

[dreamy carbon]

sorry not questioning you, but why? I thought the video specifically said to use this version

https://www.youtube.com/watch?v=McuQEzvJ-rk&t=0s from here

[ornate elk]

this video is 11 months old xD in AI time thats years, a lot changed

[vocal burrow]

0nce you ve updated to 3.10.11, you ll have to delete your stable-diffusion-master/venv folder (it will rebuild itself using the new python)

[dreamy carbon]

That make alot of sense

so uninstall this python and get new

or just install new

[ornate elk]

you dont need to uninstall it

just upgrade it

to 3.10.11

[vocal burrow]

Please do not use python 3.11, 3,.12, etc

[dreamy carbon]

is there a cmd way to update

or just download it

[vocal burrow]

download

[dreamy carbon]

ok

[ornate elk]

everything you need to know is here in my install guide: (also direct links to python)
#🤝｜tech-support message

[dreamy carbon]

I currently have 3.10

[ornate elk]

you have 3.10.6

[dreamy carbon]

ah

Now delete venv?

[ornate elk]

yes

and then launch the webui-user.bat again

it will then use the new python

[balmy shore]

any info on what 'selfattentionguidance integrated' and 'stylealign align' options do in sd-forge?

[dreamy carbon]

@ornate elkdoes this mean wait

It been like this for a while

nvm ok its done

and continues following the tutorial or should I follow your?

[ornate elk]

my guide is only for the installation part, if your now on the webui in browser, your done

but i would say, check for more updated guides and not videos older than 6 months

[dreamy carbon]

do you have a recommendation? I dont want to choose a video and eat sht again @ornate elk

[ornate elk]

no i dont know any recent basic videos

[dreamy carbon]

ok what if I follow this video and if theres problem I come nag you? lol

[ornate elk]

please dont follow that 11 months old video

it suggest a bad model version you dont need

also it only covers the installation, and your done with that already

[dreamy carbon]

what about lycoris and wildcard extension

[ornate elk]

not needed

[dreamy carbon]

what did it used to do?

[ornate elk]

lycoris is outdated and got integrated into lora

[dreamy carbon]

so get lora or

[ornate elk]

wildcard can still be used but your new, so dont xD start easy

[ornate elk]

get a model (checkpoint)

[ornate elk]

if you watch that video, you can follow up from minute 6:04

that should be fine

[dreamy carbon]

ty

[ornate elk]

np but get the dreamshaper v8 version

its the latest version

[dreamy carbon]

ok ty

[ornate elk]

at minute 17, you shouldnt do that (setting ESDN) , and hes wrong, it doesnt improve anything

[dreamy carbon]

alright

[ornate elk]

also i would recommend try out the werbui for yourself instead of copy images and reproduce them

you will learn more with just trying out

[dreamy carbon]

Thats what im thinking but I need to learn prompts and neg prompt

[ornate elk]

for negative you can try these basics:
disfigured, kitsch, ugly, oversaturated, grain, low-res, Deformed, blurry, bad anatomy, disfigured, poorly drawn face, mutation, mutated, extra limb, ugly, poorly drawn hands, missing limb, blurry, floating limbs, disconnected limbs, malformed hands, blur, out of focus, long neck, long body, ugly, disgusting, poorly drawn, childish, mutilated, , mangled, old, surreal, text, blurry, b&w, monochrome, conjoined twins, multiple heads, extra legs, extra arms, fashion photos (collage:1.25), meme, deformed, elongated, twisted fingers, strabismus, closed eyes, blurred, watermark

[dreamy carbon]

ty

[fair oxide]

*If possible, it would be easier to notice if you write in a reply.

[vocal burrow]

oh, my bad. actually for training, finetunning related question. You might be better off asking in #🔧｜finetune

[fair oxide]

Oh… Thank you, I understand.
Copy and paste it over there.

[grim ermine]

my lora files arent working and im not sure what is the problem because i only get error messages for some but none for others

[ornate elk]

whats the error you get?

[grim ermine]

RuntimeError: The size of tensor a (32) must match the size of tensor b (64) at non-singleton dimension 3

im not sure if thats the whole error

[ornate elk]

make sure the loras you try are for the spcific model version

sdxl and 1.5 need different loras

[grim ermine]

yes they all say that their base model is sd 1.5

which is what im running

am i wrong?

[ornate elk]

can you screenshot the full error and your used settings?

[grim ermine]

i cant get a screenshot of the error but i think this is the whole error

[ornate elk]

is your webui updated?

[grim ermine]

i mean i fresh installed it today

how can i check if it needs an update?

[ornate elk]

no its okay, if you installed it today

can you share me the link to the model and lora ?

[grim ermine]

ye sure give me a second

i dm'd it to you

[bronze moon]

hi

where are the bots on this server?

i dont have a gpu but i love stable diffusion where have they gone ;-;

[sudden tinsel]

Hi all,

I'm trying to install Automatic1111 on a reinstalled PC, and I'm having troubles to install torch:

Installing torch and torchvision
Looking in indexes: https://pypi.org/simple, https://download.pytorch.org/whl/cu121
Collecting torch==2.1.2
Using cached https://download.pytorch.org/whl/cu121/torch-2.1.2%2Bcu121-cp310-cp310-win_amd64.whl (2473.9 MB)
ERROR: THESE PACKAGES DO NOT MATCH THE HASHES FROM THE REQUIREMENTS FILE. If you have updated the package versions, please update the hashes. Otherwise, examine the package contents carefully; someone may have tampered with them.
torch==2.1.2 from https://download.pytorch.org/whl/cu121/torch-2.1.2%2Bcu121-cp310-cp310-win_amd64.whl#sha256=9925143dece0e63c5404a72d59eb668ef78795418e96b576f94d75dcea6030b9:
Expected sha256 9925143dece0e63c5404a72d59eb668ef78795418e96b576f94d75dcea6030b9
Got 8bcfe0b97168c713c041994fcc1774031d22ca72daf9d84796425b268a81aa2d
I tried a manual installation with pip:

pip install torch==2.1.2 torchvision==0.16.2 torchaudio==2.1.2 --index-url https://download.pytorch.org/whl/cu121
with the same result

I tried to instal the 2.1.1 version and it was sucessful so I suspect the isuue is in the original repo.

Do you know any other way to install torch 2.1.2?

[errant mauve]

whats the best image tagger tool to implement in comfyui workflow?

[vapid sphinx]

Can anyone help me by explaining how the supermerger extension works (the Merge Block Weights function and the LoRA merging)?

[coral bridge]

Can anybody help me with this problem, the image mophes constantly. Im with a 1111 web ui

[indigo ivy]

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

I think controlnet is causing this issue because i've never had this happen prior to it

[vocal burrow]

run pip cache purge to get rid of your corrupted torch package and try again

[vocal burrow]

#1047610792226340935

[snow valve]

Is there a version of https://github.com/Stability-AI/generative-models/blob/main/scripts/sampling/simple_video_sample.py that works with 24 GB VRAM or less?

I don't want to use diffusers because it hides too much of the detail and I've been bitten by not being able to repeat my environment nicely when huggingface is down. (further if I get around to training a finetune, I'd like to be able to handle just keeping track of the relevant safetensors file rather than the full huggingface repo deal)

[frosty radish]

sorry to ask but python 3.10.9 work with controlnet because i dont have preview ?

[ornate elk]

Yes should work

Do you have the models

[frosty radish]

yes

[ornate elk]

Did you clicked the explosion icon?

[frosty radish]

???

wait

lol so it was that ... 1 day lost ...lol thanks

i m running on a old laptop with a gtx1060 ti 6g of vram (wort 700$) ... it amazing it work locally

[indigo ivy]

RuntimeError: don't know how to restore data location of torch.storage.UntypedStorage (tagged with privateuseone:0)
I think controlnet is causing this issue because i've never had this happen prior to it

[ornate elk]

Nice, make sure to use --xformers --medvram --no-half-vae for the best performance

[indigo ivy]

happens when I try to use highres fix

[frosty radish]

ok thanks i ll add your arg 🙂

[frosty radish]

yes it work with your arg thanks

[ornate elk]

Nice np 🙂

[noble belfry]

which is more important, that height and width are divisible by 64 or that the resolution is? (assuming using standard aspect ratios, nothing wacky)

[frosty radish]

a other one ? ok it show me my skeleton full body like this

but when i edit the skeleton it give me this ...no face ?

it try to add one but it ask me a json file ?

[noble belfry]

its mostly for adding expressions from other images

[frosty radish]

ok i dont have any 😦

[noble belfry]

make one with controlnet, press the json button it will save it

[frosty radish]

it try , it dont save the face ?

???

i ll do a other test ..i ll be back

[frosty radish]

when i save the skelton i got no face ?

i got wholebody.py

[silver vault]

hi, i was wondering if anyone could help me with a quick question. i just got a1111 and when i generate images, it says it runs out of VRAM at 3GB. But my computer has 8GB of Vram and isn't using anything else. is there any way i can allocate more to the program so it stops failing?

[fair oxide]

Hey everyone, new here and following the pinned AMD tutorial, when installing the HIP SDK I'm getting error 215, any idea why? My graphics card is a Radeon 6600xt

(clicking get more details gets me here)

[stoic bobcat]

umm help I reinstall window and when I try reinstall SD at first I use python 3.12 which doesn't work so I uninstall it and install 3.10 now it got this how do I fix it?

it still seems to want to use old path

to wrong version

I even deleted the 312 from path here

ya know what I just reinstall the whole thing and it just work so yea

[ornate elk]

Hey install the latest Adrenalin driver

Then install the hip SDK

[ornate elk]

Hey, checkout my guide at Step 5 for the vram optimisation you need:
#🤝｜tech-support message

[ornate elk]

Deleting the venv would have fixed it

[stoic bobcat]

thanks will try to remember that next time similar problem arise

[silver vault]

thanks! just to be clear, it should look like this?

[ornate elk]

Yes

[fair oxide]

It appears I'm getting errors even trying to update my adrelanin drivers, might update to w11 while we're at it and clean c drive

[frosty radish]

ok i figure a way out ... adding a face is not easy you need a view where you can see all the dot a there proper place open mouth is better to move the dots. thanks to you

[ornate elk]

you can update but win11 is not needed for SD.
but in any case open up a cmd as admin and run
sfc /scannow
that will check windows for os file corruptions

[fair oxide]

Oh thanks, it said that it did find some corrupt files and repaired them

Lets try updating drivers again

[ornate elk]

run windows update and restart the pc

before trying

[fair oxide]

ohh alr

[fair oxide]

Seems like I'm getting another error on windows update 😅

"Retry" gives the same error after too 🤔

Should I just try and format the PC? Idk if there's any quick way to fix that

[ornate elk]

nah thats a common error fro ma few weeks ago

[vocal burrow]

there's a windows 10 update that has some problems with some computers for the past few weeks

I had it too. Let it be for now, there's probably gonna be another update to replace this one in a couple of weeks

[fair oxide]

Oh I see

Then I should just reboot and try installing the drivers right? (On this alr)

Should I try to uninstall "AMD Software" and try installing drivers again?

[golden fern]

Um hello im new in python also in stable difussion but I already searching the solution of my eror in here but nothing works, I watching this guy video
https://youtu.be/onmqbI5XPH8?si=uLDZdhrTTD0Rh_de

um maybe there's any new updates turtorial?

Btw this is my trouble

C\Unstable-diffusion\stable-diffusion-webui>git pull
Already up to date.

exit code: 1

stderr:
C:\msys64\ucrt64\bin\python.exe: No module named pip

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

[ornate elk]

dont install the pro driver

the blue one

[fair oxide]

Okay! Installed the right one this time

Should I remove the HIP SDK stuff since it gave an error last time?

And try to re-install it? Or just go with it

[ornate elk]

no you can leave them as is

[fair oxide]

Alrighty, I'm gonna keep following the pinned tutorial let's see if I can complete it ><

Thanks for the help ^^

[vocal burrow]

use the pinned tutorials please, this video is almost old enough to drive.

[golden fern]

Okayy

[upbeat gyro]

Hello, today i wanted to train an stable diffusion model, but, When i tried to extract the stable diffusion model into the working folder for dreambooth, It gave me some error regarding githubcontent, for some reason i cant extract my models when yesterday it seemed fine?

[karmic crown]

Note to self: do not upgrade your Mac!

It looks like, for some strange reason, some samplers are not working correctly on Macs updated to Sonoma 14.4 🤯

[upbeat gyro]

I use windows, but if i ever get a macbook ill keep it in mind, thanks

This basically happend if i try to train an stable diffusion model, please help!

[karmic crown]

socket.gaierror: [Errno 11004] getaddrinfo failed

those are connection problem, address cannot be resolved

requests.exceptions.ConnectionError: HTTPSConnectionPool(host='raw.githubusercontent.com', port=443): Max retries exceeded with url: /CompVis/stable-diffusion/main/configs/stable-diffusion/v1-inference.yaml (Caused by NameResolutionError("<urllib3.connection.HTTPSConnection object at 0x0000025A8C2FB940>: Failed to resolve 'raw.githubusercontent.com' ([Errno 11004] getaddrinfo failed)"))

[upbeat gyro]

I already thought that, but i find it odd how theres a connection problem, 2 days ago everything worked fine, even the extension list wont load

[karmic crown]

connection to raw.githubusercontent.com fails

i can try to help with python and webui issues, but i cant help here, sorry

[upbeat gyro]

Thats okay, maybe this site is down and it will probaly work tommorow or another day, i guess ill try tommorow, thanks 😄

[vocal burrow]

mac updated your python ?

[karmic crown]

No, Mac OS update (14.4) breaks the samplers!

i am still on 14.3.1

and it works fine for me

[vocal burrow]

yup connection issues, ditch wifi for a cable. depending of your region get a better dns, or turn it off. Also why using auto1111 1.8.0 RC ?

[noble belfry]

im suprised apple let you update. they like to make mobile devices obsolete and useless

[karmic crown]

https://github.com/AUTOMATIC1111/stable-diffusion-webui/discussions/5461#discussioncomment-8806920

[upbeat gyro]

I used kohya before to train an re:zero checkpoint, but it showed a black image if i generated, and to me, dreambooth ui for stable diffusion was the best option

[karmic crown]

APPLE MacBook Pro 14 M3 Pro 18GB/1TB
purchased this month 🙂

[vocal burrow]

as much as I love to take a dump on apple that would be the worst move on their part to turn off updates this early :p

[vocal burrow]

there was probably errors in your kohya_ss setup then

but I was more questioning the why sticking to the RC version ?

did you simply not rename the folder ?

[upbeat gyro]

Too lazy, to me, image gen is image gen, i dont care much about the name since the image generator is what matters 🙂

[upbeat gyro]

There weren't, if i train, no errors show up if it saves the checkpoint, thats the weird thing, it always shows black images for sm reason

[karmic crown]

i am installing MacOS 14.4 in VM, so I can test this myself

[ancient plover]

What webui auto 11 extension is best for connecting to civitAI? I recently got civitai browser+. I connected the api key and everything, and It worked at first, but now i cannot search for anything in webui. getting, KeyError: 'totalPages', but was getting a different error before restarting. Any help with settings would be appreciated, or should I use a different one?

[karmic crown]

api key in civitai browser plus?

for "normal" use to download models, you do not need that

i have never set it

[ancient plover]

hmm, kk ill remove it

[karmic crown]

Feature: Filter option to show favorited models. (requires personal API key)

Useful for CivitAI supporters, you can use your own API Key to allow downloading Early Access models

[ancient plover]

Im still getting this error:
File "C:\stable-diffusion\stable-diffusion-webui-directml\extensions\sd-civitai-browser-plus\scripts\civitai_api.py", line 535, in update_model_list
(hasPrev, hasNext, current_page, total_pages) = pagecontrol(gl.json_data)
File "C:\stable-diffusion\stable-diffusion-webui-directml\extensions\sd-civitai-browser-plus\scripts\civitai_api.py", line 482, in pagecontrol
total_pages = f"{json_data['metadata']['totalPages']}"
KeyError: 'totalPages'

[karmic crown]

have you tried to update extensions?

[ancient plover]

is there anything i should change?

[karmic crown]

looks fine, try to update the extension
i can't check my settings, since i am running test instance of forge in vm 🙈

@vocal burrow i cant make a1111 work in vm, but i managed to run forge and i am getting normal images

[ancient plover]

nevermind, extension broke from Civitai Api change as per git, 2 hours ago

[vocal burrow]

I mean I can go back to my parent's place and try on their apple 2. No not an m2, an apple 2. Surely one letter in the name won't matter that much.
Jokes aside it looks like the sampler did absolutely nothing. It's just noise.
The lack of log does not help figuring out what could be going wrong,

did you try to run a x/y script output ?

[karmic crown]

no

just a normal generation

[vocal burrow]

I don't know why it would cause it. But maybe ?

[karmic crown]

and in cpu only mode

i cant make it work with gpu in vm 😦

i even tried to mess with memory limit

[zinc arrow]

i have a question, is there a way i can locally host a chatbot and connect it to comfyui/any other ui, and have it generate images for me while also being able to chat with it, like bingAI, but locally on my pc without limitations

sorry for the ping but ik u guys might probably have an answer @ornate elk @vocal burrow

[zinc arrow]

like a pygmalion model connected to stable diffusion

[vocal burrow]

There are many solutions for that, eg : https://github.com/AUTOMATIC1111/stable-diffusion-webui/pull/9708/files , https://github.com/hallatore/stable-diffusion-webui-chatgpt-utilities , etc
Not sure why you d want to use them unless you re generating hundreds of images like a some kind of factory. If you re putting the effort to ask chatgpt what you want.... then just put it directly into a1111.... plus none of those solutions are gonna know about the lora/embeddings/regional prompter/controlnet, etc

There is no "read my mind cause even I don`t know what I want exactly myself" extension yet

@zinc arrow

[zinc arrow]

Nono i mean like a normal chatbot using a pygmalionai model, but like if i tell it to generate an image, it actually outputs one

[vocal burrow]

I don't think so

I don't follow llm scene that much tho

[zinc arrow]

alright ty

[noble belfry]

oh, yes you can

you can have it set to trigger writing a prompt and sending it to A1111 from textgen ui and then show it inline

you can also show it a picture and it will BLIP interrogate and tell you what it "sees" 🙂

but llms locally are still kinda pointless

[vagrant tide]

I have Stable diffusion set up how it was before and it keeps generating really terrible images of just Noise I am using the same model that worked before but i dont know what i have done 😂

[vocal burrow]

that s not what they want tho

they want an all in one solution, a model capable of llm and image generation

[noble belfry]

and being a chatbot? that makes no sense. but all those things can be done

i suppose hooking it in as a chatbot to a server would be the only thing

[vocal burrow]

they want chatgpt at home basically. Probably running on a toaster with 2gb of vram too

not happening nowadays

[noble belfry]

well....6gb vram and i can run quantized 7billion models. 🙂

[vocal burrow]

sure

now add sd to that

[noble belfry]

i do

[vocal burrow]

:3

[noble belfry]

they just take turns, you can run both at once 🙂

[vocal burrow]

you re not running them both at once tho

so it takes time to switch

[noble belfry]

no, its quickly loading offloading

[vocal burrow]

depends of the hardware

[noble belfry]

not noticably really. though i havent tried hooking with forge which does a lot of loading foo

[vocal burrow]

if you only have 16gb you ll run out of place to store those models relatively fast

nvidia/cuda stream does help a lot for fast loading, not sure if there s an amd solution for that

etc

[vocal burrow]

what settings ?

I m betting that the cfg value is through the roof

[vagrant tide]

No, I have it at 7, i have tried moving it around but still bad

[vocal burrow]

ok but what settings then ?

[vagrant tide]

i solved it, Just bad model version

[vocal burrow]

you were using a lora as a sd model, weren t you.

[vagrant tide]

no

It was the correct format

Just bad version

I just went back a version on the model and it works fine now ;))

[vocal burrow]

that doesn't make much sense but ok.

[golden fern]

hello i already tried pinned turtorial but found some problems
exit code: 1

stderr:
C:\msys64\ucrt64\bin\python.exe: No module named pip

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

is there any solution please

[solemn moon]

https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/8720

[golden fern]

Thank you sir

[sage hatch]

hy guys, i have problem with IP-adapter , anybody can help? and the result of image is blurry too. I use dreamshaper XL as the checkpoint. thankss

[solemn moon]

looks like a clash in models or summin, i dunno, just guessing, maybe you can screen shot your setup

[sage hatch]

or is it because I'm wrong at placing the model in the folder?

[solemn moon]

damn bro i dunno, have to wait for someone else 😅 can you show yr control net

[viral lark]

so im really new to all of this and have been having some problems with SDLX ive downloaded the stable diffusion local UI and also the SD_XL_base and the SD_XL_refiner models form hugging face. i got all of it to download and run however the images im getting form it are underwheming to say the least. my pc is by no means a beast but i do have 8GB of VRAM i was wonding if anyone might know the problem and if there is a fix for it. ive included the image the model has put out and can give more info on the models if needed any help would be great. Cheers!

[solemn moon]

Did you download the SDXLVAE?

i think thats defos the problem here

[viral lark]

what is that? 😛

[solemn moon]

yay we found it

https://huggingface.co/stabilityai/sdxl-vae/tree/main

have to use the VAE also

[viral lark]

do i add it the same way i did for the other two models?

[solemn moon]

yea to yr vae folder

[viral lark]

oh ok i checked it and its empty so i was def missing that

[solemn moon]

and then go to settings and make sure you can set custom VAE or summin like that

[viral lark]

ok ill do download and update that then report back! thanks for the help

[solemn moon]

holding thumbs

[soft anvil]

Guys, what do I do if SDXL takes too long to load?

[solemn moon]

delete venv folder and run bat file again is my "go to"

maybe pc restart

[soft anvil]

I was generating quickly in the usual SD then I switched to SDXL there it is very slow loading do you think if I remove venv everything will work out?

[soft anvil]

already did

[solemn moon]

no idea but it will reinstall the venv folder

what do you mean by loading?

like loading the model?

[soft anvil]

I mean generating the image in txt2img

[solemn moon]

ahh yea

depends on yr setup but SDXL is slower for less fortunate setups like mine

check mine

Steps: 10, Sampler: DPM++ 2S a Karras, CFG scale: 1.5, Seed: 3892332, Size: 1024x1024
Time taken: 50 min. 37.5 sec.

bbrrrruuuhhh

[soft anvil]

Maybe it's because I have 4gb

[solemn moon]

and thats the turbo model

[solemn moon]

yea mine is too

[soft anvil]

What's your average generation time

[solemn moon]

this 😂

[soft anvil]

😅

Okay, I think it's the 4gb

[solemn moon]

yea some guy said so the other day that the 4gb'ers suffer with the XL models

[soft anvil]

So we're losers with 4gb😂

[solemn moon]

well we lose time thats for sure, but we can still participate so feeling grateful 😂

[soft anvil]

😂

[viral lark]

well hmm this was the result.... idk what happened

[solemn moon]

hmm

can you post yr setup

[viral lark]

like the ui?

[solemn moon]

yea

[solemn moon]

minimum 1024 size

i use the same sampler and find 30 steps to be good

[viral lark]

ok i set it to that having it redraw it

(it might take a min) even with 8GB of VRAM 😐

[solemn moon]

bro try find the settings where you can see models and vae on thr side

see the left side of mine

[viral lark]

what ui is that?

[solemn moon]

"quick setting"

Lobe theme

I dont have Vae cause dreamshaperXL has it already apparently

also dont see your refiner

[viral lark]

should i have dreamshaperXL? i never added it

[solemn moon]

its faster LCM model

https://civitai.com/models/112902/dreamshaper-xl

[viral lark]

im looking under the VAE settings but this is all i see

[solemn moon]

under extensions

[viral lark]

ok im downloading it now

[viral lark]

ok i got it loaded and added to the side bar

[solemn moon]

noice

i always restart completely after makijng any changes

[viral lark]

ok i restarted it

so should i be useing the sd_xl_base? or the refine model?

[solemn moon]

base on the side

then refiner in the interface

[viral lark]

i don't see the "refiner" tab is is disabled by default?

[solemn moon]

under steps

[viral lark]

oh im blind i see it now

[solemn moon]

bro its got ,me a few times lol

[viral lark]

should i go head and test it now? or are there any other steps i need to do before testing

[solemn moon]

maybe

send you setup

screen shot i mean

see if i would change anything

[viral lark]

there it is

[solemn moon]

you need prompts

size to 1024x1024

then good

[viral lark]

ok i set the size to 1024

alright ill give it a test then

[solemn moon]

kkkk

[onyx solar]

Careful, there. One less "k" and you'll need to prompt for a white hooded outfit.

[solemn moon]

literally thought that as i typed it 😅

[viral lark]

(this make take a min) will report back with result

thx so much for the help so far i would have been at this for days without it

[solemn moon]

trial and error bro
happy to help

[viral lark]

ok update i messed and forgot to add the refiner model so i got to restart

whats that

.>

[solemn moon]

control net is the genies gift

[viral lark]

oooh do tell

[solemn moon]

like you keep the design but change the reuslt

not just random images but more controlled images

[viral lark]

YES PLS

[solemn moon]

a whole world of technical headaches to go with it too

lol

not really but def worth learning

[viral lark]

is it just an add on?

or is it like a differant step model?

[solemn moon]

yup

extension i think

watch some vids on how to install etc

[viral lark]

ok i found it

ill look into adding it later once this test finishes

[solemn moon]

yea one thing at a time

[viral lark]

ETA says and hour and 30

[solemn moon]

what gpu you got?

[viral lark]

1070 gtx

with 8GB of VRAM

[solemn moon]

yea xl takes longer

[viral lark]

im hopeing it worth it tho

so far the images where not that great

[solemn moon]

loool the settings were wrong silly

but yea lets see

[viral lark]

true true

well ima go get something to eat while i wait

ill either update ya later if your here or just talk to you tommow

[solemn moon]

https://tenor.com/view/emmy-amy-poehler-peace-out-peace-sign-peace-gif-15541895

[fair oxide]

Hi. I'm training an embeddings right now. This is the first time I have done such a thing. I am using 100 images of shoes and socks on peoples feet to see if that will help with weird toes. However as I just hit 3,000 steps. I was scouting for more images, and came across another batch of really good images. My question is, can I Add more images to the embedding training, or would I have to start over with the new images? I want to proccess the images, and drop them into the same folder as the Dataset directory and reboot the training but I am unsure if doing that would currupt the proccess and make it worse.

[golden fern]

so i have tried to fix this problem from the existing ways,i found the same case but im confused how do i add the path inside webui-user.bat

My case is the same as this, only I am confused about adding paths to webui-user. bat

[solemn moon]

open bat file in notepad

add to command args

[golden fern]

allright im gonna try

FINALLY IT WORK

Thank you sir

[viral lark]

@solemn moon update if your still here... so its taking forever to render even a 680x680 size image ima have to do some research on a way to either speed it up or just find a differant model. thx again for all your help! ❤️

[smoky cape]

So I already follow quick guide for Nvidia but when I want to start I found

Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check

I have gpu 0 nvidia,
gpu 1 amd Radeon

should i use amd tutorial instead of nvidia?

Sorry my English was bad

[uneven owl]

I got a Deep Learning VM on Google Cloud, a n1-standard-8 machine with 30 GB RAM, T4 GPU. I have set up generative models repo, and use port forwarding for the Streamlit demo/sampling.py app to run in my browser. First attempt was with 15 GB RAM instance, the Streamlit sampling example from the repo got killed due to maxed out RAM. I upgraded instance to 30GB RAM, and watching top while running the demo. During downloading of the CLIP model, the process maxed out ram and got killed. Should I upgrade to even more RAM, or adjust settings?

I have seen https://github.com/Stability-AI/generative-models/issues/36, but this happens before the app is loaded.

EDIT 2: Should I just drop the sampling demos, and use ComfyUI instead?

EDIT 3: on my Windows PC with 32 GB RAM, 2080 Super GPU 8GB VRAM, low VRAM mode,the Streamlit app loading process actually peaks RAM usage on 99% for a few minutes, and the process is not killed. The app loads as expected, and I can run the default txt2img sample with 1.85 it/s.

[misty schooner]

Does anyone know what type of model this guy uses?
https://www.youtube.com/shorts/f5L-9kzmiLU

is it even SD?

[solemn atlas]

what does that mean

OutOfMemoryError: CUDA out of memory. Tried to allocate 40.50 GiB. GPU 0 has a total capacty of 11.99 GiB of which 0 bytes is free. Of the allocated memory 12.26 GiB is allocated by PyTorch, and 1.24 GiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

[ornate elk]

Hey, you need some performance arguments in your webui-user.bat

--xformers --medvram-sdxl --no-half-vae

Add them to the line COMMANDLINE_ARGS=
Then relaunch the webui-user.bat

[ornate elk]

Can you list your gpu models?

[ornate elk]

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

[golden arch]

Here we go.

Hello, I have a question. I can't install SD because of error 3, it doesn't find the path.

C:\Users\Jeremy

Creating venv in directory C:\Users\Jéremy\stable-diffusion-webui\venv using python "C:\Users\JÚremy\AppData\Local\Programs\Python\Python310\python.exe"
Unable to create venv in directory "C:\Users\Jéremy\stable-diffusion-webui\venv"

exit code: 3

stderr:
The system cannot find the path specified.

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

I’m running python 3.10.11 and enabled path and I think it's because my user folder has "strange" characters.
I was able to change the administrator name but I can't change the user folder name.
Could you help me, please?

[olive plinth]

Hi, I need some help please to fix my stable diffusion webui. I use it in local on my windows 11. I have a deprecated message about pytorch_lightning.utilities.distributed.rank_zero_only, But I don't know what I can do. In the same time I have this message to at the launch.
I don't know how solve all issues... Can you help me please ?

[ornate elk]

You need to change your user name then restart the PC

[ornate elk]

Do you need Dreambooth?
Cause thats the problem here

[olive plinth]

@ornate elk not really, I think it's used to train model, right ?

@ornate elk I would like to try to train model, to test how do that it's why I have install it, but... in fact I use model from civitai. The install of dreambooth was for a test and see how work models training

[ornate elk]

Ah okay. I Wouldn't recommend using dreambooth in Stable-diffusion-webui because it breaks it often.
Would recommend using Khoya_ss for training as its a seperate tool.

To fix your issue. Delete the dreambooth extension of the Extension folder.
Then delete the venv folder and then relaunch

[olive plinth]

it will fix that too ?

[ornate elk]

Yes

[misty schooner]

anyone knows?

[olive plinth]

@ornate elkok I m going to try it, thx to you 😉

[ornate elk]

Looks like SD, but not any special model. I guess Photon, Juggernaut or realisticvision can do that

[misty schooner]

Is it only one? or he has many merged

And I would also want to know how can he make such complex scenes

[solemn atlas]

where do i look it up?

[golden arch]

I did it! 😕

[ornate elk]

We dont know but its mostly just a good prompt. Then he upscaled it and fixed the faces maybe

[ornate elk]

Can you show me the error again now?

[ornate elk]

In the taskmanager under performance GPU

And for the webui-user.bat. eight click and edit it

[golden arch]

Creating venv in directory C:\Users\Jéremy\stable-diffusion-webui\venv using python "C:\Users\JÚremy\AppData\Local\Programs\Python\Python310\python.exe"
Unable to create venv in directory "C:\Users\Jéremy\stable-diffusion-webui\venv"

exit code: 3

stderr:
El sistema no puede encontrar la ruta especificada.

Launch unsuccessful. Exiting.
Presione una tecla para continuar . . .

[ornate elk]

Ah okay please delete the stable diffusion folder and then follow my install guide.
#🤝｜tech-support message

Important is that you need to create a new folder like in step 1

[golden arch]

Thanks!
On the other hand, the system offers me to upgrade to win 11, is it a good idea, is it worth it, could it help me to change the name of the user folder?

[ornate elk]

For now its not worth it. Win11 is still buggy

[olive plinth]

@ornate elk sorry... this error still remain during the launching without Dreambooth

After removing Dreambooth this is the launch log

[ornate elk]

What's your gpu? And is the driver updated?

[olive plinth]

I have a Nvidia RTX4080 with 551.76 drv

[ornate elk]

Is it the latest?

[olive plinth]

yes

5 march 2024

[ornate elk]

Hmm okay.
Its hard to troubleshoot stability matrix installations and I wouldn't recommend using it.
Because it just adds a layer of stuff that can break

[olive plinth]

I use this extensions :

[ornate elk]

Ohhhh

Tensor rt

Thats it

That extension causes problems

You would have to delete it too and then the venv folder again

Then the cudnn error should be gone

[olive plinth]

sorry I past new screenshot without my hotbar 😛

[ornate elk]

Okay xD

[olive plinth]

@ornate elk Tensor RT isn't usefull ?

I mean not a mandatory for RTX card ?

[ornate elk]

Its not needed. Its an experimental speedup thing for nvidia cards. But that thing isn't compatible with most of the SD stuff

[zinc arrow]

How would i do that as a person with barely any coding experience

[zinc arrow]

That’s not what i want
I also have an ryx3070 with 8gb
And what the other person said was exactly what i was looking for, a normal chatbot that automatically runs a prompt on a1111 when it hears a trigger word like generate

Not both in the same model

[olive plinth]

@ornate elk I still get this 3 errors, is it a problem or can be ignored ?

• Error caught was: No module named 'triton'
• E:\Stability Matrix\Packages\Stable Diffusion WebUI\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.
• ERROR - failed to import mmyolo - MMCV==2.1.0 is used but incompatible. Please install mmcv>=2.0.0rc4, <2.1.0.

[vestal sage]

my stable diffusion strucked with "Commit hash:" command line
how can i fix it?

[primal hawk]

what is the missing image at the bottom of SD

[untold igloo]

If I have a specific promt, generate 20 images, find 4 seeds that I like and want to rerun all 4 of them with hires. fix. Is there any way to input all 4 seeds to run them in the same go? Or do I need to generate one at the time?

[vestal sage]

help me 🥲

it's not moving forward 🥲

how can i fix this bug ?

[vocal burrow]

give it a minute

[onyx solar]

That step takes time.

[vocal burrow]

xyz script

look for it at the bottom of the txt2img in the scripts section

there you'd want to set one column.row to seeds, input your seeds and voila

[untold igloo]

Amazing, thanks

[ornate elk]

If you dont get any issues you can ignore them.
But if something is broken you may need to disable all of the Extensions and then delete the venv again.

And then reenable them one by one

[chrome rampart]

good day all and hope doing well. Does anybody know/use DeepFashion (ADetailer) ? Hugging face says This model has 5 files that have been marked as unsafe. so i dont want to just click download

[noble belfry]

The final step of having it work as a chatbot in say a discord would require programming, (or to find a repo that does links them up but i dont know of one nor have i written one).
But if you'd be happy enough having a local chatbot that you have 1-1 convos with the AI then the first step is to install text generation webui. I'd suggest looking up recent videos about its install and reading through its github for assistance, it can either be quick and painless or a weekends work. https://github.com/oobabooga/text-generation-webui that will let you do step one, run a local LLM. it already has inbuilt hooks and extensions for both working with the SD api to fetch/generate images.

[zinc arrow]

I have oobagooba and a 7b model installed

oobabooga*

[noble belfry]

oh nice

[zinc arrow]

Wait it already supports that??

and it doesnt require tokens or anything like that to link it to a discord bot right

I worded that wrong

It doesnt require payment*

[noble belfry]

no, you'd have to have someone do that part. thats the jump. but you CAN set it up to work locally and fetch from a1111's api to generate images

[zinc arrow]

Okay thats exactly what i was looking for ty

[zinc arrow]

I can use forge instead of a1111 right

[noble belfry]

yeah same thing

[zinc arrow]

Ight bet tysm

[noble belfry]

in your settings tab there should be a tickbox currently unticked called somelike like SD-pictures-api. tick it

sorry, the settings tab is called "session" in this. for some reason. anyway, tick it.

when you reload the UI that should add this menu below the text input box. final step. for that to work, you need to make sure that forge is running its API so they can talk.

to add the API is simple enough. just put --api at the end of whatever your COMMANDLINE_ARGS are in webui-user.bat in forge\webui\ heres mine:

then run forge. once its booted up you can check if the connection works by pressing that big "get list of checkpoints" button

the rest should be self explanatory. you can just naturally ask it to send you a photo or pic. depending on the temperature settings it should always send you an image.

with forge, the first image send might take a good ~1min or so but after that it'll be just as fast as if you were to type into its UI yourself.

weirdly there is a lot less lag from the original a1111.

oh just be mindful of one thing: both textgen and forge run on gradio so whichever one is open first gets port 7860, when you load the second thing it gets port 7861 so you might have to change the port number on the api

[zinc arrow]

oh shit i didnt see all that
im not gome rn but as soon as i am ill try it and get baxk to u

[noble belfry]

no problem. its all pretty self explanatory, and i think ive said everything you need, the rest is just playing around yourself. good luck!

[vital lance]

Does Stability AI REST API work in China?

[sturdy jasper]

[noble belfry]

i did a bit of looking around and haven't seen stability's name pop up in any block lists. the likes of openai bing and huggingface are however closed off

[fair oxide]

I am drawing on stabble diffusion but it is taking too long, I think my computer is good enough, can you help me why is it taking so long?
32 GB RAM
RTX 4050 6VRAM
i5-13th Gen

[ornate elk]

Checkout my guide at Step 5 for the performance increasing settings needed:
#🤝｜tech-support message

[chrome rampart]

good day all and hope doing well. Does anybody know/use DeepFashion (ADetailer) ? Hugging face says This model has 5 files that have been marked as unsafe. so i dont want to just click download

[late cove]

nahh why is it so blurry

💀

[noble belfry]

dont download non safetensors versions of files

i have an example of a model.bin which when loaded can open calc. e: its not mine, should make that clear lol.

and it is still sitting on huggingface (though put up for research purposes by someone else)

[chrome rampart]

hi, no idea what open calc is lol. Shame really like what i read about the deepfashion features but safety first. Super thks

[noble belfry]

calc.exe is your computer's calculator. its commonly used as an example in exploits to show that something malicious can run a program without you allowing it.

[chrome rampart]

so Adetailer models risk in general it looks EEK glad did not download then. Thks again really helped

[vital lance]

What about replicate?

[noble belfry]

i honestly dont know what china currently blocks as im not in china

[proven finch]

.safetensors is fine

[quiet remnant]

i just installed comfyui , after restart macbook, how can i run comfyui? please help

[zinc arrow]

does forge webui use --listen just like a1111 or do i have to add something extra

[zinc arrow]

Someone on the pygmalionAI server is telling me to use kobald, would kobald have a feature like this?

[noble belfry]

i know that having tried kobold for llms, it was - as of ~1 year ago - terrible on low end machines for memory and if IIRC you have an 8gb vram card which puts you in the low/medium end. it does pointless things like mounting itself separately and doubt it would be very efficient in terms of loading/offloading models on the fly. it might work nicer over a network but you had asked for two concurrent things locally. also that whole pygmalionAI scene isnt SFW and i'd rather not give tech support for it on this discord, questions about that might be better asked to them. 🙂

i think llms in general are out of the scope of this channel, i was just helping because there happened to be an existing tie in with SD from something i was familiar with. you could try the koboldAI discord i guess.

[zinc arrow]

Ah okay mb ty tho

[zinc arrow]

Understandable

[noble belfry]

no harm in trying with koboldAI

[zinc arrow]

Another question, actually related to sd this time,

why does --listen not work for me on forge

[zinc arrow]

I think im just gonna forget about it for now as its too much for me 😭 but ill try later tho

[noble belfry]

are you trying to run it as a server? if so you need to allow it through your router to work as a server

[zinc arrow]

Wdym as a server

I just want to be able to access it thru my phone locally

my args --listen --xformers --api

[noble belfry]

you still might need to play with your switch to allow devices on your network access each other.

[zinc arrow]

It works on a1111 but i recently switched to forge and it isnt working

[noble belfry]

oh that might be firewall. open up your firewall and allow inbound rule for local network 7860 and 7861

by default even on your network it wont let any old thing just connect

[zinc arrow]

Like this?

[noble belfry]

try it and see, you might need a UDP rule too

[zinc arrow]

--listen uses my local ip right, like 192.168.1.1

[noble belfry]

local yes

[zinc arrow]

Oh my god i was using the wrong ip this whole time im so stupuid

😭

i was using the "default gateway" instead of the ipv4 address from ipconfig /all

[noble belfry]

🙂 well now you have learned how to do this for other things too.

[zinc arrow]

it works now thankfully i was so mad i thought forge didnt have it

[zinc arrow]

yeah xd

[noble belfry]

most programs do this behind the scenes for new devices, but its handy to know a bit yourself

if nothing else you're learning networking

[zinc arrow]

yeah ive never used the firewall app before

Now i know how to add a rule xd

[noble belfry]

dont forget that those rules only apply to your machine youre on, so for anything else you need to change on your router

but for local network it should all work nice

so does it all work?

if you've set up everything correctly you should now have a local AI which when asked, will generate and send your phone an SD image

if it all works, then, for a flex, tick this. first time you use it it will install BLIP and clip interrogate an image. that means that you can take a photo on your phone and ask your local AI what it thinks of the photo

i'd just like to point out that this stuff, although janky, was introduced in open source before the big guns did demos of it.

im not sure if textgeneration webui is as mobile friendly for uploads, but still

[old parrot]

Anyone have a better picture that shows the differences between all sampling methods?

[glad fiber]

is 0.1 avr_loss good or bad?

[solemn atlas]

what is pony diffusion?

i download a lora of a character which is pony version but what do i do with that

[old parrot]

Put it in your Lora folder under models

[solemn atlas]

i did but i cant find it

the 2nd one but i cant find it on lora on sd

[noble belfry]

its an SDXL version of anime only trained models trained on only things from image boards. it has its own prompting syntax. its very good for drawings of that kind but veers towards nsfw stuff.

[solemn atlas]

thats interessting, didnt know that

but i cant use it because i can't find

[old parrot]

turn on "Always show all networks on the Lora page (otherwise, those detected as for incompatible version of Stable Diffusion will be hidden)" in settings

I had to do the same, they still work mostly

[noble belfry]

i dont know if using loras from those models work very well, in general they mangle photorealism

[solemn atlas]

where do i find this option?

[old parrot]

Settings lol

[noble belfry]

i found a silly setting the other day. if you put a soundfile into your webui root called notification.mp3 it will make that little noise when its finished

[vocal burrow]

yup there is also support for browser notifications

[noble belfry]

re our convo last night you were right about running both at the same time. BUT if you keep doing it, you CAN run a small llm and stable together with similar latency to just asking the SD client to do a thing. just as long as you never move models. 🙂

useless practically, but locally a fun toy

[golden arch]

Hi, I tried to create a new folder but

"You go into the folder you created in this case Ai, then click in the File Explorer bar (not searchbar) and type cmd then press enter."

[ornate elk]

Type only cmd

Remove the C:\ai before it

[glad fiber]

alternatively, you can open the windows menu and type cmd

[golden arch]

Thanks, the cmd part worked BUT...😥

Creating venv in directory C:\IA\stable-diffusion-webui\venv using python "C:\Users\JosÚ Antonio\AppData\Local\Programs\Python\Python310\python.exe"
Unable to create venv in directory "C:\IA\stable-diffusion-webui\venv"

exit code: 3

stderr:
El sistema no puede encontrar la ruta especificada.

Launch unsuccessful. Exiting.
Presione una tecla para continuar . . .

[glad fiber]

which step are you in?

[ornate elk]

your username has still the unsupported character

[glad fiber]

i think it should not be a problem

[golden arch]

Yes, I told you, I cant change it!

[ornate elk]

is it bound to an microsoft account maybe?
Win10 or Win11?

[golden arch]

After editing, save and launch the webui-user.bat inside the stable-diffusion-webui folder.

[golden arch]

I don't know, maybe.
I was able to change the user name but not the folder name.
win 10 pro

[glad fiber]

try to install the python to a different location and make the webui-user.bat use it

you will specify the full python path here

[glad fiber]

i dont remember how the python path is on windows, but i guess it will be something like this

[golden arch]

On C?

[golden arch]

I moved the python folder to C\ and specified the full python path, same result...

Couldn't launch python

exit code: 9009

stderr:
"C:\Python\Python310" no se reconoce como un comando interno o externo,
programa o archivo por lotes ejecutable.

Launch unsuccessful. Exiting.
Presione una tecla para continuar . . .

[glad fiber]

add the \python.exe part

i was training my lora and i was wondering why it wasnt working. The image folder was named 30_wb17myself person and if i put wb17myself in the prompt, it would show nothing. Only now I realised that if I put one of the captions I used, the output image would work correctly. How can I make my lora not to need a whole prompt to be triggered, but only a word?

[golden arch]

Did it! Is working!!!!!!😊

[noble belfry]

nevermind

[clear ice]

ight you probably heard this seven billion times but I'm trying to make a short film which will use AI video for artistic statement purposes, I am struggling to get SD running

because it either keeps giving me errors or keeps downloading things

so who's up to help me step by step

[noble belfry]

help is for specific errors, whats your first one?

[clear ice]

I will try again and show you my errors

[pale bobcat]

hello, i am following along to this video to try a train SD on a face. in the video @6:52, the guy is in the "Train Tab" and clicked on "Preprocess Images". I dont have that under my "Train" tab. I am on version v1.8.0, can someone help me. Also if anyone has a better face model trainer method, I'd love to check it out. Trying to make a hyper realistic woman

[mossy narwhal]

Any ETA on the api issues? https://stabilityai.instatus.com/

[indigo ivy]

Get this everytime I do highres
RuntimeError: don't know how to restore data location of torch.storage.UntypedStorage (tagged with privateuseone:0)

[tepid wolf]

Hey guys

Is possible to Fine-tuning the SDXL model to handle image captions exceeding the standard 75-token limit?

[ornate elk]

What settings do you use? (Resolution, steps, etc)
What's your GPU?

[indigo ivy]

GPU: RX6700xt
Resolution: 512 x 744
Steps: 25
and then with hires I was upscaling it by 1.7 with 20 steps

Ok so I ran it again with the upscaler being 4x-UltraSharp this time and it worked

Previous time I was using one of the ergans

[ornate elk]

Both should work. But make sure your using 10 hires steps and not 0

Also if you want better performance and less vram limit you should switch to the zluda version of directml

[indigo ivy]

I tried to but I think I messed up because it didnt work

one example being the SD-Zluda folder I had made somehow got deleted during the process

I followed the guide in the pinned msg btw

[ornate elk]

Yea thats my guide

It should work

No folder should get deletet lol

[indigo ivy]

I can give it another try

[ornate elk]

Sure, let me know if you have any questions

I'm on AMD too

[indigo ivy]

oh word

maybeit was the arguments?

I had mine setup like set COMMANDLINE_ARGS= --use-directml --zluda or something --medvram --no-half-vae --opt-sub-quad-attention --opt-split-attention-v1 --disable-nan-check --autolaunch

no nvm

Ill try the guide again

[ornate elk]

That won't work

Yea please follow the guide step by step

[indigo ivy]

Alright btw what gpu do u have

[ornate elk]

Im running a 7900XTX

[indigo ivy]

oh nice

[ornate elk]

But I tested a lot on 6700xt, 7800xt and vega7 too

At friends PCs

[indigo ivy]

Ah, yeah i definitely did something wrong then

[ornate elk]

Probably. I'm AFK now, (20mins) but feel free to @ me in chat, if your stuck

[indigo ivy]

Sure thing, thank you

[indigo ivy]

Hey so Im now at step 5 of the guide and Ive done the instructions in the red area but im not sure if I still have to install SD Next

[ornate elk]

No you dont need SD.next

[indigo ivy]

very nice

[indigo ivy]

So I got it set up but when i try to generate an immage I get this in the cmd
RuntimeError: invalid argument to memory_allocated

[ornate elk]

Can you show the whole error in cmd?

[indigo ivy]

yeah one sec

[indigo ivy]

Traceback (most recent call last):
File "C:\SD-Zluda\stable-diffusion-webui-directml\venv\lib\site-packages\gradio\routes.py", line 488, in run_predict
output = await app.get_blocks().process_api(
File "C:\SD-Zluda\stable-diffusion-webui-directml\venv\lib\site-packages\gradio\blocks.py", line 1431, in process_api
result = await self.call_function(
File "C:\SD-Zluda\stable-diffusion-webui-directml\venv\lib\site-packages\gradio\blocks.py", line 1103, in call_function
prediction = await anyio.to_thread.run_sync(
File "C:\SD-Zluda\stable-diffusion-webui-directml\venv\lib\site-packages\anyio\to_thread.py", line 33, in run_sync
return await get_asynclib().run_sync_in_worker_thread(
File "C:\SD-Zluda\stable-diffusion-webui-directml\venv\lib\site-packages\anyio_backends_asyncio.py", line 877, in run_sync_in_worker_thread
return await future
File "C:\SD-Zluda\stable-diffusion-webui-directml\venv\lib\site-packages\anyio_backends_asyncio.py", line 807, in run
result = context.run(func, *args)
File "C:\SD-Zluda\stable-diffusion-webui-directml\venv\lib\site-packages\gradio\utils.py", line 707, in wrapper
response = f(args, **kwargs)
File "C:\SD-Zluda\stable-diffusion-webui-directml\modules\call_queue.py", line 95, in f
mem_stats = {k: -(v//-(10241024)) for k, v in shared.mem_mon.stop().items()}
File "C:\SD-Zluda\stable-diffusion-webui-directml\modules\memmon.py", line 99, in stop
return self.read()
File "C:\SD-Zluda\stable-diffusion-webui-directml\modules\memmon.py", line 81, in read
torch_stats = torch.cuda.memory_stats(self.device)
File "C:\SD-Zluda\stable-diffusion-webui-directml\venv\lib\site-packages\torch\cuda\memory.py", line 258, in memory_stats
stats = memory_stats_as_nested_dict(device=device)
File "C:\SD-Zluda\stable-diffusion-webui-directml\venv\lib\site-packages\torch\cuda\memory.py", line 270, in memory_stats_as_nested_dict
return torch._C._cuda_memoryStats(device)

RuntimeError: invalid argument to memory_allocated

[ornate elk]

Thats Strange. Do you have any other Programm running?
Maybe restart the PC and try again

[indigo ivy]

No no other program I dont think, will restart pc

[indigo ivy]

So I restarted my pc and after running webui.bat again I notice there are errors even before I generate an img

img

[ornate elk]

Oh okay

Pls show them

[indigo ivy]

Want me to screenshot them?

[ornate elk]

Wait

Don't run webui.bat

Always run webui-user.bat

[indigo ivy]

Yeah that’s the one I initially ran but I tried webui bat too

[ornate elk]

can you show me your paths in system varaibles?

[ornate elk]

looks ok

[indigo ivy]

Yeah I’m not sure what’s going on with it

[ornate elk]

did you copied the dll files over to the torch/lib ?

[indigo ivy]

Yeah

[ornate elk]

and you have doen this correctly?

it means this path:

rename the library folder and paste the other one in there

so that you have two different folders in the rocblas folder and the new one needs to be library

[indigo ivy]

Might’ve done that wrong

[ornate elk]

after you tried it again, restart the PC

before launching the webui-user.bat

[indigo ivy]

I did do it the correct way this time and restarted my pc but now that I think about it, I forgot to run the AMD HIP SDK 5.7 file which might be the issue

[ornate elk]

which file? the installer?

[indigo ivy]

Yeah, but I just ran it and restarted my pc and it still doesn't work so i dont think that was it

[ornate elk]

you may need to delete the venv folder again and then relaunch the webui-user.bat

also make sure you replaced the librarys folder after you installed the hip sdk

[indigo ivy]

Oh alright

[river flame]

I cant use ReActor in Krita AI right?

[pallid solstice]

hi thinking about buying stable diffusion but is it just too sensitive? can it not handle jokes? gave it this prompt "digital art style image of many bankers, dressed in suits and all based on a greedy pig character and have dollars in their pockets, are running scared of one giant bitcoin in the background. This takes place on wall street." and it gave this images. tried countless times and they are always smiling rather than be scared or worried, etc. whats up with that?

i mean i love the art style it gave me but its useless cos its the opposite of what i asked emotions wise

tired a million times and have now run out of my free credits. not once it gave me the emotion i was after.

[raw kindle]

Hey guys, can somone help me with the "workflow" settings for comfyui ? I managed to install SD XL Turbo, but i try all this samplers, cfg number, steps and things, but i dont get good pictures.

i want to get pictures like in this free sdxl turbo online page: https://stablediffusion.fr/xl-turbo

[vernal cargo]

anyone got any tips for using anything v3 on stable diffusion? trying to recreate something like this

[frozen tapir]

what could be a possible reason for my backend showing error, when using a refiner. I can't use refiner to generate a image

[indigo ivy]

yep it works now

Thanks for all the help man

[ornate elk]

Perfekt np

[vernal cargo]

anyone know some good models that can produce an image of this quality???

[ornate elk]

Which webui?

[ornate elk]

divinieanimemix

[karmic crown]

MacOS 14.4 definitely fu...d Euler A, for some unknown reason 🤯 fortunately new samplers Euler A SGMUniform and Euler A Turbo in forge work without any problem

[river pumice]

where do i put my own upscaling models in auto1111 sd?

[opaque jungle]

Can I bug you for a sec? I'm getting the following during first image gen:

RuntimeError: invalid argument to memory_allocated

[karmic crown]

post more details, copy text from cmd or make a screenshot

[opaque jungle]

---
Traceback (most recent call last):
  File "G:\Stable Diffusion\stable-diffusion-webui-directml\venv\lib\site-packages\gradio\routes.py", line 488, in run_predict
    output = await app.get_blocks().process_api(
  File "G:\Stable Diffusion\stable-diffusion-webui-directml\venv\lib\site-packages\gradio\blocks.py", line 1431, in process_api
    result = await self.call_function(
  File "G:\Stable Diffusion\stable-diffusion-webui-directml\venv\lib\site-packages\gradio\blocks.py", line 1103, in call_function
    prediction = await anyio.to_thread.run_sync(
  File "G:\Stable Diffusion\stable-diffusion-webui-directml\venv\lib\site-packages\anyio\to_thread.py", line 33, in run_sync
    return await get_asynclib().run_sync_in_worker_thread(
  File "G:\Stable Diffusion\stable-diffusion-webui-directml\venv\lib\site-packages\anyio\_backends\_asyncio.py", line 877, in run_sync_in_worker_thread
    return await future
  File "G:\Stable Diffusion\stable-diffusion-webui-directml\venv\lib\site-packages\anyio\_backends\_asyncio.py", line 807, in run
    result = context.run(func, *args)
  File "G:\Stable Diffusion\stable-diffusion-webui-directml\venv\lib\site-packages\gradio\utils.py", line 707, in wrapper
    response = f(*args, **kwargs)
  File "G:\Stable Diffusion\stable-diffusion-webui-directml\modules\call_queue.py", line 95, in f
    mem_stats = {k: -(v//-(1024*1024)) for k, v in shared.mem_mon.stop().items()}
  File "G:\Stable Diffusion\stable-diffusion-webui-directml\modules\memmon.py", line 99, in stop
    return self.read()
  File "G:\Stable Diffusion\stable-diffusion-webui-directml\modules\memmon.py", line 81, in read
    torch_stats = torch.cuda.memory_stats(self.device)
  File "G:\Stable Diffusion\stable-diffusion-webui-directml\venv\lib\site-packages\torch\cuda\memory.py", line 258, in memory_stats
    stats = memory_stats_as_nested_dict(device=device)
  File "G:\Stable Diffusion\stable-diffusion-webui-directml\venv\lib\site-packages\torch\cuda\memory.py", line 270, in memory_stats_as_nested_dict
    return torch._C._cuda_memoryStats(device)
RuntimeError: invalid argument to memory_allocated

[ornate elk]

whats your gpu?

[opaque jungle]

Got the RX 6800 XT

Oh I'm seeing some more errors above the one I posted:

    RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu! (when checking argument for argument index in method wrapper_CUDA__index_select)

[ornate elk]

did you do every step correctly?

show me your paths in sys variables

[opaque jungle]

Oh- SYS variables

I think I added them to the user variables

[ornate elk]

oh

[opaque jungle]

Let me try adding to both

[ornate elk]

remove them from users

only add them in sys

[clear ice]

First error

[ornate elk]

whats your gpu?

which guide did you followed

[clear ice]

i have intergrated

AMD

[opaque jungle]

Will try now. Should I remove venv or anything or try starting the ui?

[ornate elk]

yes, delete the venv folder and relauch thr webui-user.bat

[clear ice]

CPU is Ryzen 7 5700G

[ornate elk]

okay, what settings did you tried?

[upbeat gyro]

Hello, what are the best settings to train an stable diffusion checkpoint for anime? It keeps getting messed up faces

[clear ice]

I just tried running this, I'm currently tinkering and trying to get it, I'm trying to generate videos based on image but make them look clearly AI generated for a personal project, I was trying to merge 2 images

just installed FYI

I may need models

[ornate elk]

you need to use models that are 2gb

and not larger

then dont use a too high resolution

[ornate elk]

did it worked?

[opaque jungle]

I'm not getting an error now- however image generation doesn't work and not getting an error or anything. Sec.

just stays in queue. Not sure if it's doing anything or not

[ornate elk]

the first image takes 15-40 minutes

[opaque jungle]

Shouldn't there be a progress bar of sorts?

I'll let it sit like this for a while

[ornate elk]

no

not for the first gen

[opaque jungle]

Alright thanks!

[ornate elk]

if the generate button is grayed out its fine

just wait

[opaque jungle]

yeah, looks good

[ornate elk]

cmd will show nothing for a long time