🤝｜tech-support

[ornate elk]

I didnt make a guide for CPU mode but you only need to use these args in the webui-user.bat
--use-cpu all --skip-torch-cuda-test --no-half --precision full --opt-sub-quad-attention

AMD igpus like Vega can use DirectML but with its amount of vram you can't do much with it.
You could allocate more RAM to the GPU in BIOS, then it could be a better experience

[kind tapir]

That's cool you can do that in the BIOS, I did take the compTIA certs so I get what you mean

I added those args, no idea if it was supposed to work after that or not but

[ornate elk]

Can you show how you added it?

[kind tapir]

notepad

[ornate elk]

Looks correct

Try delete the venv folder and relaunch

[kind tapir]

oh the path is still in quotation marks

will that work?

[ornate elk]

Yep

[midnight roost]

is it possible to use supermerger for merging loras to checkpoints with amd gpu?

[ornate elk]

Should work, but didnt tried it

[midnight roost]

under device its on cuda should i change it to cpu? also the error i see when i try is
ModuleNotFoundError: No module named 'scripts.A1111'

[keen raptor]

i have a question

rtx 4060 is 8gb and 3060 is 12gb, so like, for image generation purposes is 3060 better since it has more Vram?

[kind tapir]

hmm, asks me for admin permission to delete, I click continue (I am admin, bad practise I know)

It continues to recycle about 40,000 files and then says 'you need permission from (this device\my account) to make changes to this folder

Options are try again, and cancel

:/

[graceful adder]

any reason you don't just add the python binaries to your PATH?

[kind tapir]

I'm not exactly sure what that means, but when I reinstalled python (the compatible version) earlier I did check the box saying to add python to PATH

[graceful adder]

if you open up powershell and type python --version do you get anything?

[graceful adder]

yeah so you shouldn't need to specify your python location since it's available to anything that runs, since it's in your path

[ornate elk]

Because you installed it not the correct way and also you still have it in the downloads folder

But the venv will be recreated at launch

[ornate elk]

Yes a 3060 with 12gb could do more vram demanding tasks like using larger models

[keen raptor]

hmm, do you think 4060 would be enough for SD anyways

8gb

[ornate elk]

Yea its okay for 1.5, sdxl and pony models and even can run flux or sd3 with their smaller model versions

[keen raptor]

o

i jus wan make sum good 2048x2048 images

not more than dat

[ornate elk]

That will work with upscaling

[karmic crown]

do not generate that directly, first make smaller images and then upscale

@ornate elk check your messages 🤣

[keen raptor]

ikikik

i meant

is it capable of making cool stuff

like

i forgot the name

‘deforum’

[vale sparrow]

I have a laptop with 6gb vram that i use at work and I see sd3.5 medium is only 5gb, will that run in the 6gb vram?

ah.... it needs a 9gb clip file lol

[ornate elk]

you can use the smaller clip file t5 q8 gguf

[vale sparrow]

I gave it a test as i had already started downloading the files and it does indeed work on my 6gb vram laptop... I dont understand how... but the default workflow is working

takes about 1 minute per generation though

looks like its unloading the model each time too

i suspect the clip is using the cpu not the gpu

oddly the gpu does not seem to be used other than as a vram storage

[iron lantern]

where can i get help with stable ai iam getting grey images ?

also getting this when i generate " NansException: A tensor with NaNs was produced in Unet. Use --disable-nan-check commandline argument to disable this check. "

[zealous willow]

IA image is what needs the most VRAM if possible.
I can confirm Generative fill exist with Stable Diffusion. Especially with control net + llama (see post from eliyaviel), but I supose youtube videos describe it. (not as good as photoshop though because it only fills one 4 directions at least if there was no update since)
If you can get your hand or any "RTX" card that would be good.
Most AI is better running on "cuda" technology.
I personally was able to install a1111 (or forge webui or sd next) without removing my latets python.
You dont need to specify it there, instead I changed it in: webui.bat
I changed "if not defined.. and straight put my python".
Another info: The first python path you found was the "shortcut" to find the real one, you should have right clicked it -> properties -> target. That's the path.
Another note: I can feel you know computer science a bit but out of practice for a moment, because you are not aware of chatgpt and such capabilities. It's simple: You can ask chatgpt (by showing it the code in webui.bat and ask how to modify the code to always point to a constant python path for example) 🙂
You are in the right track.

[late monolith]

Hi. First time trying to train my own model with Dreambooth. Gives me the following error: "AttributeError: module 'transformers.integrations' has no attribute 'deepspeed'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "F:-=AI\Git\stable-diffusion-webui\extensions\sd_dreambooth_extension\dreambooth\ui_functions.py", line 734, in start_training
from dreambooth.train_dreambooth import main # noqa
File "F:-=AI\Git\stable-diffusion-webui\extensions\sd_dreambooth_extension\dreambooth\train_dreambooth.py", line 34, in <module>
from diffusers.training_utils import unet_lora_state_dict
File "F:-=AI\Git\stable-diffusion-webui\venv\lib\site-packages\diffusers\training_utils.py", line 27, in <module>
if transformers.integrations.deepspeed.is_deepspeed_zero3_enabled():
AttributeError: module 'transformers.integrations' has no attribute 'deepspeed'
Duration: 00:00:40
Duration: 00:00:43"

Anybody got an idea what to do?

[nocturne lance]

does anyone know if the 3.5 large good for 3060?

[glossy compass]

has anyone used 6950xt for the FLUX pro?

is there a never version of FLUX or a better model?

[ornate elk]

Thats a model problem.
Which one did you tried?
Is it larger than 2gb?

[ornate elk]

For AMD its the best to use the flux gguf models

[glossy compass]

how longs the gen time

[ornate elk]

Idk I guess around 1-2 mins for one image

For my 7900xtx, flux dev takes 50-60 seconds

[lost crescent]

Has anybody got OmniGen working in SD Next at all?!

[finite zenith]

Hi

How do I change clip skip for stable diffusion forge?

I tried using 'all' in UI setting but changing the clip skip from there doesn't do anything

Still getting the same image with different clip skips

Nevermind just found out the difference is very subtle for SDXL models

[ornate elk]

Crypto scam wow

[fierce moat]

Flux pro is open sourced?

[ornate elk]

Nope

[hallow latch]

Question.. I am rendering some images with a Pony style and I see that it uses more of my PC RAM and very little on my GPU. Why?

The GPU went up maybe to like 20% as max?

[ornate elk]

What are your txt2img settings?

[hallow latch]

My command line is
set COMMANDLINE_ARGS=--no-half-vae --medvram-sdxl --theme=dark --xformers

[late monolith]

Is there anybody that can explain the "Exception training model: 'module 'transformers.integrations' has no attribute 'deepspeed''." for a noob like me?

[ruby yoke]

what were you doing when you got that error?

[iron lantern]

anyone know how i can add two bodys or girls in the one image or its very hard to do ???

still learning how to add the models and loras

1st image is where the models go right ???
2nd image is where the loras go also ??? any help is grateful

[karmic crown]

looks good

[karmic crown]

with most anime models those should work 1girl, 2girls, 3girls or 1boy, 2boys, 3boys

[hollow rune]

hi everyone, even using the default workflow, some time the output become all black, but if I queue another image, it output image normally, why is this?

[lethal canopy]

why can't i download

i did everything right from what i know

[ornate elk]

You have the wrong Python Version installed

You need 3.10.11 64bit

[lethal canopy]

were can i find that

[iron lantern]

my loras are not showing up in the drop down menu now

[karmic crown]

loras should be the same type as model, for example 1.5 lora and 1.5 model, or xl and xl to be shown

there is an option to show all always, but i do not have a1111 installed at this moment

@ornate elk please send a screenshot where that is if you can

[ornate elk]

https://www.python.org/downloads/release/python-31011/

[lethal canopy]

i installed but this

[karmic crown]

remove 3.13

[lethal canopy]

i did

[karmic crown]

oh, wait

remove venv folder

[lethal canopy]

i went to Local\Programs\Python and manually delted it

[lethal canopy]

were is what venv

[karmic crown]

this one

just remove it

it will be created again

[ornate elk]

You need to delete the venv folder

[lethal canopy]

ok did i do it right

@ornate elk

[karmic crown]

looks good enough

[iron lantern]

Do I just put this in the promt like 2 girls etc

[karmic crown]

yes

2girls the rest of your prompt

2girls, blue hair, red eyes, green eyes, red hair, long hair, day, long sleeves, witch hat

for example

[iron lantern]

Allright my lora are gone after putting in my new models into my folder

Put in the perfect pony xl model now all my inside out lora are not showing up in stable diffusion 😡

They are in the lora folder ... maybe just reload the Web browser?

[spiral comet]

I have question, is that normal I have like 50 sec for render in flux with 1280x1280? GF 4080 (16gb).

[karmic crown]

you need the same type of model and lora

you cannot use incompatible loras

you can show them by force, but there is no guaranty that it will work

[ruby yoke]

use SD 3.5 large and prompt for them

Prompt: we are standing at the front of a stage, looking up at the actors. on the left side of the stage, We see Elon Musk dressed as george washington. He is pointing at iron man. Iron man is standing on the right side of the stage, with his hands on his hips. And behind both of them, we see a blond anime girl dressed as a housekeeper. A chair and a table are placed near the back of the stage, with dark green curtains hanging down behind them.

[lucid summit]

Is there an image resize node that expands the image instead of stretching it? I swear i've seen it somewhere before

[ruby yoke]

@limber pier might know

[limber pier]

best you can really do is composite it onto a blank one if that's what you mean

[late monolith]

Trying to train a model with my own pictures

[iron lantern]

@karmic crown so i got the loras and the models do i ijust add the promts now

[lucid summit]

Found it actually

https://github.com/cubiq/ComfyUI_essentials it's the ImageResize+ node part of this repo

[karmic crown]

you need to combine lora triger with your prompt

and if you have lora:mrs_Andersen:1 be sure that your lora is actually is mrs_Andersen.safetensors for example
they should be the same

if your file is called for example mrs_Andersen_something.safetensor, than it will not work

but you should be fine, based on what you posted so far

[iron lantern]

ok so can i just copy paste what the creator of the lora has posted for promts to see if it works ??

[karmic crown]

that is always the best aproach

[iron lantern]

also how to add the preview image in the lora tab

its not coming up

[karmic crown]

put a picute with the same name in lora folder

something.safetensors
something.jpg

or
something.png

[iron lantern]

i tried to genarate the image and got this text in the bottom of the image

[karmic crown]

like this

[iron lantern]

its driving me up the bloody wall this

[karmic crown]

install xformers as it says

my battery will die

[iron lantern]

where do i get that

[ruby yoke]

you should probably just install SwarmUI and let it do all the technical work for you, and then just generate in it with ComfyUI

[iron lantern]

i just spent 3 days trying to figure this out lol

[vocal burrow]

edit webui-user.bat with notepad, find the set COMMANDLINE_ARGS= line and add --xformers to it. relaunch webui-user.bat, voilà.

[craggy epoch]

clean install on windows.. do you still do python 3.10.6 or go for 3.10.15 (latest)?

automatic1111 or comfy or forge being the end results

or is there a better version of python to use? (I've seen some stuff about 3.11 being fine, but wasn't sure).

[iron lantern]

it just get bad to worse wtf

[vocal burrow]

Any 3.10 python (newer to or 3.10.6) should work fine.

[vocal burrow]

no guarantees that 3.11 python will work, it might but it might also break as soon as someone/some extension calls a python function with a new prototype.

[vocal burrow]

what s your gpu ? what s the complete log (copy paste everything from the cmd into a .txt and drop said .txt file in here) ?

[iron lantern]

its a 1650

ok gimme a sec

@vocal burrow

this is when i load a model or try to generate

[vocal burrow]

ok add --no-half to the command line args too

and --medvram

[iron lantern]

i did and when i add medvram it wont work

[vocal burrow]

what do you mean it won t work ?

[vocal burrow]

which webui are you using ?

[iron lantern]

iam not sure

i think its pyhton

[vocal burrow]

no

but ok

where did you get that sd.webui folder

[ornate elk]

Its missing the = ?

Can you show the webui-user.bat again?

[gaunt lodge]

I changed the model folder name for my forge and now it doesnt recognize where the files are at. Both models and Loras

[ornate elk]

Yea dont change the model folder names

Revert it and it should work again

[iron lantern]

I have to go out do it later thanks for helping

[timid zodiac]

Ok this may sound super stupid, but I was training dream booth with some images and it was going for like 40 minutes and now the power for the entire street is out, there isn't a chance this is my fault right

[iron lantern]

I just downloaded it

Really driving me crazy thinking of giving up on this

[vocal burrow]

but where ?

[iron lantern]

I followed this guide

It is nsfw I don't want to get banned

Can delete it if it's not aloud

I followed this and it ended up not working

Real angry as I just wanna make images

[vocal burrow]

Removing it just to be safe as it s really explicit.

I kept the link and will check it out later on to see what you installed. But I won t be able to check it out properly until tomorow (at work+party tonight)

In the meantime, I'd suggest doing a proper clean install using the guide in this channel's pinned messages.

[iron lantern]

Ok where abouts

I'll just remove the whole folder and start again

Does the guide in the channel do nsfw ?

[vocal burrow]

there is no filters for local instances

[iron lantern]

Ok where is this pinned thing y,ou said

[vocal burrow]

click the pin logo

[iron lantern]

Iam on mobile lol

Atm

[iron lantern]

Does this mean no filter when I run it locally on my pc

[vocal burrow]

yes

[iron lantern]

Ok thanks ... I'll have a look at the guide ... when I get back on my laptop .... also do you recommend that I delete the whole folder and start over ?

[vocal burrow]

cant check the content of your guide right now cause I m at work.

so I can t evaluate how much of a mess your install is (or is not)

Plus as I said I won t be able to check it until tomorrow

So a complete reinstall is probably the faster and easier way

shouldnt take more than an hour even if it s your first time.

[hallow latch]

I'm having some issue with auto 1111 breaking for me at random times. I have been doing like 6-10 promts with 20 images a piece without a problem and next prompt it breaks and it even have happend after a second prompt. I have enough memory .. could anyone help me with diagnostics?

[iron lantern]

Ok ill just delete all of it and start again

[marble rampart]

Hi, can I ask something about my account?

[vocal burrow]

what s your gpu ?

[vocal burrow]

depends which account you re talking about.

[iron lantern]

Is my laptop the problem or it's just a messy install ?

[hallow latch]

I have a GTX 3060 with 12 GB ram

[marble rampart]

2 days ago I subscribed the $19 package but I can't login to my account, it keeps show "login" button but never shows that I'm logged in.

[upper sky]

Hello. I don't know if I'm in the right place, but I have a question. I am currently using Automatic1111 on SD 1.5 but I am having problems creating texts. I'm still in the learning process, I would appreciate if anyone can give me some tips and tricks. I know how checkpoints and lora works, I have no clue on VAE and others. Thank you for your time

[iron lantern]

Also have this problem

[vocal burrow]

what command line arguments are you using ?

[vocal burrow]

but which website ?

[vocal burrow]

should be fine once you ve reinstalled and followed instructions from the pinned guide

[iron lantern]

I have a laptop with 8gb and a 1650gtx is this ok

[vocal burrow]

VAE quick guide :
1/ What is a VAE.
It's a part of the stable-diffusion pipeline that encode/decode information from tokens to latent space and from latent spaces to pixels. Aka it transforms math statistics in pictures
2/ Where do I put my VAE ?

• VAE with .vae.pt, .vae.ckpt, .vae.safetensors extensions go into the models\Stable-diffusion folder
• VAE with .pt, .ckpt, .safetensors go into models\VAE
  3/ How do I use my VAE ? Three possibilities :
• Either you name it similar to another one of your model (eg : Anything-V3.0.safetensors + Anything-V3.0.vae.pt), by doing that it should automatically load the VAE when you load the associated model.
• You manually load your VAE by going to Settings -> Stable-Diffusion -> sd_vae and selecting your VAE
• You add an easily accessible VAE dropdown at the top of your page to quickly switch back VAE by adding sd_vae to your Settings -> User Interface -> Quicksettings list

[iron lantern]

Iam in Canada with my fiancee atm so I did not bring my pc

My pc has 16gb with a 970gtx

[vocal burrow]

both could run SD but it will be slow

[iron lantern]

I don't mind waiting as long as it will work

[upper sky]

Thank you, this helps a lot!
Now all I have to learn is how to create texts, if it is possible. Cheers mate!

[hallow latch]

COMMANDLINE_ARGS=--no-half-vae --medvram-sdxl --theme=dark --xformers

[vocal burrow]

should be pretty easy thanks to the t5 text encoder that comes with SD3 architecture. pretty much out of the box

For older architectures, there are some workflow out there you can google but it will require more knowledge, extensions and tinkering

[vocal burrow]

should be fine with that, then you re probably asking too much of SD, would need you to screenshot the webpage to see which settings you re using when generating.

[upper sky]

I'm still on SD 1.5 but I will look over SD3. Thank for your help! Have a blessed day!

[vocal burrow]

yeah with sd1.5 you re probably gonna want to add the text yourself and then do a img2img pass on it so it can be "redraw" with the text in it. Or you can try to use controlnet (canny or edge or depth) with an image of the text you want to inprint.

or etc etc.... there are probably a bunch of other way to approach this issue.

[marble rampart]

https://stability.ai/stable-assistant

[vocal burrow]

Ok so good news it s not a fake website. Unfortunately we pretty much are volunteers in here so we don t have access to accounts or administration stuff for this website. I d suggest you to get in contact with their support https://stabilityplatform.freshdesk.com/support/home

[hallow latch]

Yesterday .. I maybe did 8-12 before it accured..

[ornate elk]

Set the hires steps to 10

[hallow latch]

When I was going to bed I put 6 in que and it only did 1=20 images + 2 on second que.

[vocal burrow]

the whole webpage please (if it s nsfw, change it)

[hallow latch]

Better?

[ornate elk]

You can also install the tiled VAE extension that can prevent out of vram errors

Its very helpful when used with hires fix

[vocal burrow]

still incomplete, i can t see what s above the prompt or the footer.

also what s inside the adetailer section

(also, yes hires steps set to 0 means "do as many hires steps as native res steps" which is often overkill)

[hallow latch]

Above the prompt is just the header?

[hallow latch]

Is it something that needs to be enabled everytime or is it in the background running?

[ornate elk]

Its a checkbox like adetailer

Has to be checked once for the generations

[hallow latch]

I run default adetailer, never changed anything in there

[ornate elk]

Until a restart

[hallow latch]

I'll download it and see if it works tonight 🙂

[vocal burrow]

not sure it s very effective without a prompt for it

[vocal burrow]

even above, checkboxes for the model and depending of your configuration, vae, clip skip, etc

screenshoting the whole page would have been easier

[vocal burrow]

https://media.discordapp.net/attachments/1002602742667280404/1111097170150563881/image.png?width=810&height=228 (after you installed the multidiffusion-upscaler-for-automatic1111 extension)

just tick Enable TiledVAE

leave everything else as default

[hallow latch]

I just use it for the face fixes, never knew you needed a prompt for it?

[ornate elk]

it doesnt need a prompt

but it can help for specific stuff like getting red eyes etc

[upper sky]

One last question and I am out of your way, does SD3 works only in ComfyUI?

Or I can use the basic, uh...stable diffusion 😄

[ornate elk]

SD 3.5 yes, but SD3 works in auto and forge too

[upper sky]

Perfect, thanks. ComfyUI isn't for me, it gives me a headache, haha.

[ornate elk]

yea forge will be the next webui with SD3.5 support

and i guess SDnext has the support too

[vocal burrow]

without a prompt it s just gonna redraw without instructions soooo results will be kinda random.
A generic prompt such as "beautiful detailed face" should cover 99% of use cases.

[upper sky]

My PC isn't that powerfull, I can't even use Flux Dev, I lack the VRAM

[hallow latch]

Ohh cool thanks ❤️ I'll look up adetailer prompts etc and I will install VAE extenson tonight

[iron lantern]

So just delete the whole file and reinstall everything there is no reason why it won't work ..... it's just because I installed it wrong ?

[vocal burrow]

probably yes

stop at the first sign of errors and report it there.

[iron lantern]

At first I was getting greyed out images when I generated them and then I started screwing around and now it's a big mess

[vocal burrow]

probably because you have a 16xx gpu

follow the instructions and it will be fine

[marble rampart]

Thanks for suggestions, unfortunately they did not provide me the answers I needed. Maybe I have to change my bank account instead of letting them deduct my fees after the trial.

[iron lantern]

So what u saying is my gpu cant support it ??

[vocal burrow]

not what I m saying.

it works but requires some extra steps ( explained in the guide )

[iron lantern]

Ohh would my pc be better it has more ram but a 970 gtx

[vocal burrow]

Probably not, ram does not matter that much apart from the loading model part of the pipeline.

the important part is the gpu (and its ammount of vram)

[ornate ruin]

Can anyone give me a simple step by step of exactly how to get consistent faces via Face IP? I have the control Net installed, and have chosen the batch option to insert several images. In my case I have chosen 6 images of my self. 3 selfies 2 portraits and a full body portrait. But that alone isn't enough it seems

I use Realistic Vision 5.1 and Ultrasharpx4 upscaler...
Thanks 🙏🏼😅

[iron lantern]

So my 1650 is not the best but it will work it came with my laptop lol

[snow swallow]

Hello, can anyone point me to the tutorial on installing stable diffusion on a machine with ryzen 5 4550g ?

[ornate elk]

hey, does the machine has no dedicated gpu ?

[snow swallow]

That is correct. I only have: AMD Ryzen 5 PRO 4650G with Radeon Graphics 3.70 GHz and 32gb of ram on win10

[ornate ruin]

Hey CS1o 😄 maybe you can help me once again haha. Maybe you even somewhat remember my system and the problems I was facing the last time. 😅😂

[ornate elk]

Radeon TM Graphics ?

[snow swallow]

I'm not sure what that means. Is there a way I can check it?

[ornate elk]

can you screenshot the gpu in taskmanager

under performance

[ornate ruin]

Not trying to Help leech, I'm sorry. But I genuinely can't get it to work by myself, I'm super sorry

[ornate elk]

hey, whats the issue

[ornate ruin]

Still trying to get the consistent faces

But failing lol

[ornate elk]

you have two options.
Either use the slow cpu mode. or use the amd directml mode which is very limited with 4gb vram but faster

[iron lantern]

So just delete everything and follow the guide here ?

[snow swallow]

limited in what sense? Thanks for your answers. Is there a tutorial with the step by step instalation? I'm a newbie

[vocal burrow]

they re not answering you

[vocal burrow]

you just follow the pinned guide and it will be fine

[iron lantern]

Allright not meaning help hog

Just annoyed as I have been trying this for 5 days now

[ornate elk]

in the pinned messages are the guides for installing it.
CPU mode is currently missing

limited in a way you cant generate larger resolutions

[snow swallow]

Last time I experimented with SD a lot of times my machine would freeze. Is there a diference regarding this issue in either using directml or slow cpu mode?

[white stirrup]

hello i get this error when using the example 3.5 medium workflow from huggingface (all components were fresh downloads)

[ornate elk]

that depends on the RAM of the machine too. both cpu and directml mode can freeze the pc

[white stirrup]

running on a 10gb 3080

which is supposed to be suported properly

latest drivers

[white stirrup]

@ornate elk ?

[ornate elk]

idk, sd 3.5 support is very fresh so maybe a bug

[white stirrup]

what is the issue if you dont mind telling me

is it just the model thinking it cant run on the card?

[ornate elk]

size mismatch could be an incompatible lora, vae or other stuff uses with the model

[white stirrup]

the vae seems to come from the model and everything else is downloaded straight off of hugingface for 3.5M

[ornate elk]

hmm then idk, havent tried 3.5 yet

[white stirrup]

probably a bug

because 3.5L worked

and 3.5M is supposed to be a cut down more consumer model

[hallow latch]

Is it the middle one here?

[ornate elk]

yes

[formal trout]

I'm trying a workflow and got this

It is not installed with "missing custom nodes"

[iron lantern]

wtf am i doing wrong i followed the guide here

[karmic crown]

have you checked this one while installing?

[iron lantern]

iam not sure

do i have to reinstall again i guess ?

[karmic crown]

reinstall and be sure to check it

whatever it says on yours

[iron lantern]

ok i left it unchecked

ill check it now reinstalling

is this ok i have a 1650 card

[ruby yoke]

did you update comfy first?

[white stirrup]

Uh

No i updated like a few weeks ago

[ruby yoke]

comfy released updates the day sd3.5 medium was released to support a new function one of the workflows for it has. you need to update with the update.py file in the updates folder located inside the comfyUI folder, not with manager

[white stirrup]

Ah thanks

Thats the issue then

[white stirrup]

I use comfy manager, that should also work right?

[ruby yoke]

probably, yes. comfy posted about it on his annouce channel in his discord

[white stirrup]

Dm me the link?

[ruby yoke]

usually not. not for comfy itself. if you've changed anything about the default, inlcuding where you have models, that can fail. use the update batch file in the updates directory inside the main comfyUI directory

[white stirrup]

Alright

[karmic crown]

medvram

you have n instead of m

[ruby yoke]

DM'd you the invite link to comfy's discord

[white stirrup]

Ah i didnt see

Gotta love message requests

[iron lantern]

so remove the -- ?

[karmic crown]

no

[iron lantern]

ohhh

[karmic crown]

just change N to M

[iron lantern]

vram

[karmic crown]

at the end

[iron lantern]

my bad

[karmic crown]

you have
-- medvran

it should be
-- medvram

[iron lantern]

got it

so just let it do its thing now ?

do i still download the models like perfect pony xl now ?

[karmic crown]

looks good

[iron lantern]

and my loras while i wait ?

[karmic crown]

it will be faster if you do not download anything else

since it need to download some things

[iron lantern]

what model can i use thats anime style

i have my loras that i wanna use

what models do i need to look for ??? perfect pony models that will work with my software ?

[karmic crown]

i never used pony

i use lightning

[iron lantern]

what model will work with what iam downloading

any links would be great ... wanna go for that uwu anime style

is this what iam looking for

ofc the model is nsfw iam not gonna post the image

[karmic crown]

be sure that your loras and chekcpoint (model) are of the same type

https://civitai.com/models

for example

i like lightning since it requires only 8 steps

[iron lantern]

ok so if iam using a pony model

i need pony loras

sorry for being so dumb rather learn this now

was gonna pay someone to do these images for me

[karmic crown]

just be sure not to mix loras and models

it would most likely fail

so pony and pony

[iron lantern]

ill dm u the link iam using since its nsfw @karmic crown

just to double check before i start this headache all over again

check dms is that ok

[hallow latch]

Should I be worried about this?

[fair oxide]

Yes @hallow latch

[hallow latch]

How can I fix it?

[fair oxide]

Sent

[hallow latch]

Sorry, I am not accepting a discord invite to another server when I can get help here.

Offical support desk of what?

[fair oxide]

For guide and to resolve issues

[hallow latch]

I would not call it an Offical support Desk .. That is more of an unoffical help server where you probably aim to help people with the people in the server.

[fair oxide]

You see

[hallow latch]

Yeah I don't think so, this is a tech support and this is more of a offical discord server for stable diffusion.

[fair oxide]

It's

[ornate elk]

Try update the extensions

[hallow latch]

I did it before today when I downloaded the VAE thing :/

[ornate elk]

Hmm okay

Maybe check the issues on dynamic prompts github

[karmic crown]

works fine for me

[magic lintel]

what on the earth does this mean?

ValueError: could not convert string to float: '1.1.2'

[karmic crown]

you have 1.1.2 which can not be converted to decimal number

you need 1.12

[iron lantern]

and it starts again

WTF IS THIS

[karmic crown]

--xformers

you probably have only one -

[iron lantern]

ok ok

[karmic crown]

and fix that RAN alredy

it should be RAM

medvram

🙂

M, not N 🙂

as CS1o would like to say
show my what you have in your webui-user.bat

[karmic crown]

that is good

[magic lintel]

thanks, i was being an idiot, forgot to say so earlier

[iron lantern]

i give the fuck up

iam fucking done with this bullshit

day 8 of this crap now

[ruby yoke]

now will you try SwarmUI?

[iron lantern]

Sorry I give up with this bullshit it's turned me off all this

It's annoyed me to the point now when I don't want to anymore it just won't work foe me

[thin crescent]

I installed zulda version on 7800xt as per pinned guide @ornate elk. when im trying to use sdxl model getting error. it always end with Failed to create model quickly; will retry using slow method.

[iron lantern]

last thing i will ask for help if this is not gonna work after this then iam done

wtf is this shit i followed all the steps 8 days of trouble shooting for all this crap now iam getting real mad

[ruby yoke]

the reason i said install swarmUI is because it will take care of all the stuff youv'e been fighting with, automatically, and you won't have to deal with it. to quote someone else "swarm just... works" - plus you've been trying to configure one of the more technically challenging interfaces out there.

so if you install SwarmUI, let IT deal with all the technical stuff, and then run ComfyUI inside it, you can just concentrate on generating images, not trying to turn into a software devloper/system administrator

[warm warren]

does anyone know how to fix this issue?

[warm warren]

ok thanks

[main rune]

@ornate elk Hey I've been meaning to ask you if you know how to work the zluda setup into Stability Matrix

[balmy shore]

after doing git clone to initially download the repo before like running or anything, is it okay to move that downloaded directory? i assumed yes and brought it up by one directory but i realized i never asked anyone and am not sure

also i cant get xformers working, i asked chatgpt how and all it did was tell me how to use git to download or install it, but that just broke my webui, and after reinstalling torch and torchvision i still cant get xformers working, but the webui is working
generation is using cuda but not xformers, it might not even besomething i need but if anyone knows lemme kno

oh, but back to my original question, i feel like moving the folder initially before running anything is somehow causing these issues or related to them, im not sure

[vast pulsar]

so i just got

forge

what args should i always have?

[ornate elk]

That error is currently normal, after that it should still work normaly.
If you only have 16gb of RAM not vram you need to increase the Windows Pagefile.

[ornate elk]

Nope, stability matrix is just an installer for the webuis. The easiest option would be to install the webui with my Guide and then replace the one installed with stability matrix

[ornate elk]

You can try --cuda-stream

[ornate elk]

Hey, yes moving the folder after git cloning breaks the git path. So you can't update the webui for example.
Its recommended to always git clone directly into the folder where the webui should stay.
And to install xformers you only need to add --xformers to the webui-user.bat

[balmy shore]

oh heck

if i recreate the folder name that i originally git cloned into and move it back there would that fix it?

F:\stable-diffusion-webui-ux\stable-diffusion-webui-ux
is where i git cloned, and then i moved it all up one directory cuz i didnt like the fact that it was duplicated

ok it didnt dang

if i git clone again int he same dir will it fix it

[balmy shore]

or should i just delete the whole dir and start over, after backing up my models and such

[iron lantern]

Any clue what's causing this or how I can figure it out

[ornate elk]

Yea better start over

Then you can be sure everything is correct

Also what's your GPU?

You have a lot of wrong commandline_args in the webui-user.bat

[balmy shore]

gtx 1060

oh

[ornate elk]

A GTX 1060 only needs --xformers --medvram --no-half-vae

Never add --skip-torch-cuda-test

[ornate elk]

You have the folder two times maybe cloned into each other. So better delete everything and clone it one time

Then edit the webui-user.bat with the commands above and your fine

[balmy shore]

okie im moving my models out now and gonna delete and do that

i use --listen because thats the one that lets me login to the ui from my phone over wifi, afaik

and the insecure extension one was required for an extension i installed at one poitn but i cant remember which so ill leave that out for now

ill get it working with your recommended ones before messing with shit for now

[ornate elk]

Yea --listen is for local access in your wifi

But dont enable the insecure extension access.
Better install all needed extensions before adding --listen

Because --listen blocks installing extensions

[balmy shore]

o i didnt know that

[keen raptor]

Ello

Erm

What's the difference between installing Sd 1.5 and Sdxl, i mean, is there a difference if I upload sdxl models in sd1.5

[balmy shore]

most webui's seem capable of running both or either, xl is a larger newer architecture on a different base model than 1.5

xl needs more vram and stuff so the difference is that xl might be able to generate larger images at the cost of needing more memory or handling larger prompts or something i havent used xl models so im not sure ,

awesome it installed xformers correctly this time.

im using this one btw https://github.com/anapnoe/stable-diffusion-webui-ux-forge

[next eagle]

Hello, I need help setting up kohya_ss from this tutorial:
https://www.run.ai/guides/generative-ai/stable-diffusion-training

Something with the python versions I have make it incompatible with the kohya_ss setup.bat

Reinstalled the proper versions as prompted to, still get version errors

[ornate elk]

Delete the venv folder in kohya ans relaunch

Then it should use the updated python version

[next eagle]

I did this already

Let me show you my error real quick

[fossil crag]

Hello Everyone,

I'm happy to be here and hopefully I'll get my questions answered.
I'm using Stability AI API for my image generation service called Francis AI. I'd like to make a subscription / buy credits as a company but I can't find any form / input that makes it possible, nor in the stripe checkout nor on the UI of the Stability AI client.

If anyone knows or faced this issue before, I'd really appreciate some further information 🙇‍♂️

Thank you guys in advance and sorry for disturbing peace with my silly question

[brittle lantern]

No idk where to ask this my fren has a M2 pro and 16gb of ram like I do on his his MacBook and I was told GGUF models was good to use so I recommended it to him but he doesn’t know which Flux dev GGUF models to get for his specs since it goes from Q2.K to Q8

So anyone have a idea on which model would perform the best on his specs of MacBook Pro

[karmic crown]

i am mac user, but unfortunately i cant answer that, since i haven't tried that yet
i have only used 1.5, XL, Turbo and Lightning

but...

unlike windows, ARM macs have unified RAM, which means it is used for both CPU and GPU

in case of 16gb it would be like having around 10GB GPU

so, my advice would be to use models that should work on 8GB on windows

also, which command line params he has in webui-user.sh?

https://github.com/viking1304/a1111-setup/discussions/2

[next eagle]

@ornate elk please help with thois

[karmic crown]

you need to downgrade your python

[next eagle]

I already did

[karmic crown]

ERROR The python version needs to be greater or equal to 3.10.9 and less than 3.11.0
and you have 3.11.9

[next eagle]

it says that

[karmic crown]

if it use venv, remove that venv and create it again

[next eagle]

but when I uninstall and install the right version, then stable diffusion doesn't work and also kohya_ss setup.bat

it's weird

Ok I will try this

gimme a moment

[karmic crown]

which UI do you use for stable diffusion?

comfy? a1111?

[brittle lantern]

i do not know

[karmic crown]

i posted a few different command line combinations, and when they should be used

[noble belfry]

does it look like noodles or a big form?

[ruby yoke]

if you're using comfy, it uses an embedded version of python rather than your system version.

[next eagle]

I don't know

[noble belfry]

screenshot?

[next eagle]

I just followed the guide that I posted

[next eagle]

Of what?

[noble belfry]

whatever youre using

[next eagle]

How do I do this?

[noble belfry]

phone?

[ruby yoke]

how do you take a screen shot?

[noble belfry]

if on computer, its a button to the right of F12 top of your keyboard should say prnt scrn

[ruby yoke]

@next eagle there are a number of different interfaces you can run stable diffusion in, we need to know which one you're using

[next eagle]

I am uninstalling every python I have and deleting the Venv folder as well and after that open web_ui.bat

[ruby yoke]

windows key+shift key+s on windows 11

[ornate elk]

Stable works good with 3.10.11 and kohya too

[next eagle]

Can I stream it for someone?

[ruby yoke]

1. don't use webUI 2. install SwarmUI and let it do all the technical stuff and just run ComfyUI inside it

[next eagle]

Is there a tutorial for this please?

[noble belfry]

left or right

[next eagle]

The goal is to train my own Ai

[next eagle]

Right

[ornate elk]

Just install python 3.10.11 and uninstall 3.11 and then install kohya

[noble belfry]

ok you're using a1111

[next eagle]

I've done this 4 times

doesn't work

[noble belfry]

scroll down to the window there should be really small writing, could you copy that and paste that?

[ornate elk]

Open up a CMD and type
Python --version

[next eagle]

The right version does not workk

[noble belfry]

ok, so its not installed, thats cool. listen to what @ornate elk says now

[next eagle]

ok

ty

[noble belfry]

(just a universal rule for other people, even if someone else is typing, listen to CS1o :D)

dont worry jeff cq, we've all been where you are, youre going into space 😄

[ruby yoke]

yeah - @ornate elk is lead tech. pay attention to him

[next eagle]

thanks

It's weird I just uninstalled it and it shows I have it installed

Both microsoft and website downloads

[noble belfry]

the S stands for SAINT, due to patience 😉

[next eagle]

This one is stilled installed but it's from microsoft

[ruby yoke]

rule #1 - if you uninstall something REBOOT!

[next eagle]

Oh gotcha

[ruby yoke]

and do that before you do anything else. clear the sytem out and load everythign fresh

[next eagle]

gotcha

thank you

[ornate elk]

Yep after the restart delete khoya and git clone it again into the folder

Then it should work

[ruby yoke]

and supertech

[karmic crown]

ok, sell windows computer, buy mac, and use my script
or listen to cs1o and aryetis when it came to windows 🙂

[next eagle]

lol

[noble belfry]

i'd honestly go latter.

[karmic crown]

mac is not fast enough for some things (related to SD)
but for some basic things it is ok

[noble belfry]

yeah but if the answer to "my dog is sick" is "get a fish" its not really an answer

i cant get into my emails, because of MS office....
linux tech support nobody wanted in batman voice: get linux, i might not be the OS you want but im th-

"no its ok, i got them in the recycle bin"

[karmic crown]

when i am serious, my answer is always:
if you already have a mac, just use it
if you have windows use that one

and if you want to buy a new computer just for SD, buy windows (its much cheaper)

[noble belfry]

i know this isnt a social channel but thats an oddly french answer

[ruby yoke]

it is if you're tired of vet bills for the dog

the real answer is 'you're trying to do something that's very technical, but you don't have the skill set for it. it's time to learn those skills'

[noble belfry]

the real answer people want is just how to do thing they were in the middle of trying to do. like they might have got home, said to friends, gather round i am doing a thing now. then thing didnt happen. the answer they need is never "UR HOUSE SUKS LOL" "LERN TO BILD" 😄 just let people continue their thing, especially with text to image which is very train of thought. a lot of people are now thinking (ok i know most are thinking good big girl very yes) but a lot of people are looking up artistic terms and techniques now, i think its lovely

[safe inlet]

here i am prompting for "painted with a brick" and getting the same results

[iron lantern]

how do i find 1.5 loras ?

i found out that my pc can only do 1.5 and wanted to find jill anderson from inside out models but not sure where i can get a 1.5 model for her

[hallow latch]

You can go to https://civitai.com/

[iron lantern]

if i wanted to find rileys mom from inside out 2 in 1.5 how do i do that ?

[hallow latch]

Make sure you select like this

then go to filters to the right side and take these

That way you get checkpoints and SD 1.5 ( don't take the wildcards ) one hahah 😄

Here you can search all different checkpoints and loras etc.. As you can see @iron lantern

You can DM me if you have any further questions or ask them here and maybe someone else will respond as well 🙂

[hallow latch]

Forgot to answer but search for Lora's if so.. You have a checkpoint model I guess that you like. Then search for Riley mom or whatever and select Loras and SD 1.5

[iron lantern]

check your dms 🙂 @hallow latch

[hallow latch]

I have a question regarding Tiled VAE .. Currently doing SD Upscale with in enabled and running Encoder Tile Size: 1536 and Decoder on 96 and I keep getting this .. Should I turn it off during SD upscale or should I change the values?

[ruby yoke]

Unfortunately, what they're trying to do is frame a house and they don't know how to actually hold a hammer correctly yet. and you can't successfully frame a house till you learn the basics of how to hammer a nail.

[balmy berry]

Ok, frustration setting in here, tried to install video abilities in my A1111 setup, and now this is all i get when I try to start up... File "E:\sd.webui\webui\launch.py", line 48, in <module>
main()
File "E:\sd.webui\webui\launch.py", line 44, in main
start()
File "E:\sd.webui\webui\modules\launch_utils.py", line 465, in start
import webui
File "E:\sd.webui\webui\webui.py", line 13, in <module>
initialize.imports()
File "E:\sd.webui\webui\modules\initialize.py", line 23, in imports
import gradio # noqa: F401
File "E:\sd.webui\system\python\lib\site-packages\gradio_init_.py", line 3, in <module>
import gradio.components as components
File "E:\sd.webui\system\python\lib\site-packages\gradio\components_init_.py", line 1, in <module>
from gradio.components.annotated_image import AnnotatedImage
File "E:\sd.webui\system\python\lib\site-packages\gradio\components\annotated_image.py", line 8, in <module>
from gradio_client.documentation import document, set_documentation_group
File "E:\sd.webui\system\python\lib\site-packages\gradio_client_init_.py", line 1, in <module>
from gradio_client.client import Client
File "E:\sd.webui\system\python\lib\site-packages\gradio_client\client.py", line 24, in <module>
from huggingface_hub import CommitOperationAdd, SpaceHardware, SpaceStage
File "E:\sd.webui\system\python\lib\site-packages\huggingface_hub_init_.py", line 487, in getattr
submod = importlib.import_module(submod_path)
File "importlib_init_.py", line 126, in import_module
File "E:\sd.webui\system\python\lib\site-packages\huggingface_hub\hf_api.py", line 47, in <module>
from tqdm.auto import tqdm as base_tqdm
ModuleNotFoundError: No module named 'tqdm.auto'

Might anyone have any thoughts on how to get unstuck from this hell that I seem to be in? I can't start SD, as long as this error keeps coming up... 😑

[ornate elk]

you have to install the webui the correct way by following the install guide from the pinnes messages

then it will work

[balmy berry]

so, I need to RE-install the webui...

[ruby yoke]

after you uninstall the mess you have

[balmy berry]

I will be VERY unhappy if I lose all the extensions and models I have installed...

[hallow latch]

Make a backup of them..

Thats why I do

[balmy berry]

Ok, thank you all, guess I'm gonna be busy for a bit... 😕

[ornate elk]

you can move them over

[ornate elk]

np, also which video extension did you tried to install?
some are outdated and wont work

[hallow latch]

Can you recommend a extension that work to make ai videos? 😄

[ornate elk]

for 5-10 second videos you can test out AnimateDIff

[balmy berry]

aw, heck, I don't even remember anymore. I haven't slept since yesterday morning trying to figure this out... Probably should do THAT first thing... 😆

[ornate elk]

then dont copy over all extensions, you could break the webui again

[balmy berry]

Think one was mov2mov, and the video extras may have been the other, I deleted em though, after the break.

[ornate elk]

install them one by one

mov2mov can work with a workaround that includes replacing a file

[hallow latch]

I'll look into, any settings to must have or just roll with it? 😄

[ornate elk]

it needs motion module files. and important is to set its batch size lower if you have less vram than 24gb.
default is 16, better test out 8 or 12

[balmy berry]

I just wanna make some videos, damnit! lol

[ruby yoke]

then use mage.space and stop worrying about local

[balmy berry]

kk, thanks, it's definately nap time now. My cat is looking at me like "yo, we gonna cuddle sometime this week, ya ass?"

[ruby yoke]

and animate with kling or luma

[balmy berry]

all of which cost how much? (fixed income, disabled vet... no extra money for much of anything that isn't "free")

[hallow latch]

So it not enough to just download AnimateDiff?

[fossil crag]

Hello Everyone,

I'm happy to be here and hopefully I'll get my questions answered.
I'm using Stability AI API for my image generation service called Francis AI. I'd like to make a subscription / buy credits as a company but I can't find any form / input that makes it possible, nor in the stripe checkout nor on the UI of the Stability AI client.

If anyone knows or faced this issue before, I'd really appreciate some further information 🙇‍♂️

Thank you guys in advance and sorry for disturbing peace with my silly question

[ornate elk]

you need the motion modules from civitai too
https://civitai.com/models/108836/animatediff-motion-modules

[hallow latch]

Thats for 1.0 .. I am using mostly 1.5 so I need 1 for 1.5 then right?

I am not misstaken right? 😄

[ornate elk]

there are all listed

choose the v2 for 1.5

[hallow latch]

Ahh aight ❤️ Thanks

[safe inlet]

there is a v3

[hallow latch]

Damn, I remember I had problem with AnimateDiff before .. it just crash auto 1111 and don't remember what I had to do to fix it :S

[fossil crag]

anyone knows anything about stability AI api? I'd like to pay for it with my business deatils but I can't find a way to do it

[hallow latch]

I just get

[ruby yoke]

https://platform.stability.ai/docs/api-reference

https://platform.stability.ai/pricing

[ornate elk]

whith naimeatdiff?

[hallow latch]

Yeah .. I had this problem before with it .. Not sure what setting I had to run with to make it work but it was pain in the ass :/

[ornate elk]

make sure to lower the batch size value

[hallow latch]

Default is 10?

[ornate elk]

idk i thought 16

try 8 maybe if oyu 12gb vram

[fossil crag]

I currently have an application that is using stability AI but I couldn't find any B2B option to provide my tax detail, do you have a more specific link or something?

[ruby yoke]

you will have to contact them through the contact form on that website. that's the only way to get in touch with them for business related questions

[fast owl]

Trying to install a1111, getting an error code and im not really of a programming background to figure it out

  File "C:\Users\adria\a1111\stable-diffusion-webui\launch.py", line 48, in <module>
    main()
  File "C:\Users\adria\a1111\stable-diffusion-webui\launch.py", line 39, in main
    prepare_environment()
  File "C:\Users\adria\a1111\stable-diffusion-webui\modules\launch_utils.py", line 423, in prepare_environment
    run_pip(f"install -r \"{requirements_file}\"", "requirements")
  File "C:\Users\adria\a1111\stable-diffusion-webui\modules\launch_utils.py", line 144, in run_pip
    return run(f'"{python}" -m pip {command} --prefer-binary{index_url_line}', desc=f"Installing {desc}", errdesc=f"Couldn't install {desc}", live=live)
  File "C:\Users\adria\a1111\stable-diffusion-webui\modules\launch_utils.py", line 116, in run
    raise RuntimeError("\n".join(error_bits))
RuntimeError: Couldn't install requirements.
Command: "C:\Users\adria\a1111\stable-diffusion-webui\venv\Scripts\python.exe" -m pip install -r "requirements_versions.txt" --prefer-binary
Error code: 1
stdout: Collecting setuptools==69.5.1 (from -r requirements_versions.txt (line 1))
  Using cached setuptools-69.5.1-py3-none-any.whl.metadata (6.2 kB)
Collecting GitPython==3.1.32 (from -r requirements_versions.txt (line 2))
  Using cached GitPython-3.1.32-py3-none-any.whl.metadata (10.0 kB)
Collecting Pillow==9.5.0 (from -r requirements_versions.txt (line 3))
  Using cached Pillow-9.5.0-cp310-cp310-win_amd64.whl.metadata (9.7 kB)
Collecting accelerate==0.21.0 (from -r requirements_versions.txt (line 4))
  Using cached accelerate-0.21.0-py3-none-any.whl.metadata (17 kB)
Collecting blendmodes==2022 (from -r requirements_versions.txt (line 5))
  Using cached blendmodes-2022-py3-none-any.whl.metadata (12 kB)
Collecting clean-fid==0.1.35 (from -r requirements_versions.txt (line 6))
  Using cached clean_fid-0.1.35-py3-none-any.whl.metadata (36 kB)
Collecting diskcache==5.6.3 (from -r requirements_versions.txt (line 7))
  Using cached diskcache-5.6.3-py3-none-any.whl.metadata (20 kB)
Collecting einops==0.4.1 (from -r requirements_versions.txt (line 8))
  Using cached einops-0.4.1-py3-none-any.whl.metadata (10 kB)
Collecting facexlib==0.3.0 (from -r requirements_versions.txt (line 9))
  Using cached facexlib-0.3.0-py3-none-any.whl.metadata (4.6 kB)
Collecting fastapi==0.94.0 (from -r requirements_versions.txt (line 10))
  Using cached fastapi-0.94.0-py3-none-any.whl.metadata (25 kB)
Collecting gradio==3.41.2 (from -r requirements_versions.txt (line 11))
  Using cached gradio-3.41.2-py3-none-any.whl.metadata (17 kB)
Collecting httpcore==0.15 (from -r requirements_versions.txt (line 12))
  Using cached httpcore-0.15.0-py3-none-any.whl.metadata (15 kB)
Collecting inflection==0.5.1 (from -r requirements_versions.txt (line 13))
  Using cached inflection-0.5.1-py2.py3-none-any.whl.metadata (1.7 kB)
Collecting jsonmerge==1.8.0 (from -r requirements_versions.txt (line 14))
  Using cached jsonmerge-1.8.0.tar.gz (26 kB)
  Installing build dependencies: started
  Installing build dependencies: finished with status 'error'

stderr:   error: subprocess-exited-with-error

  pip subprocess to install build dependencies did not run successfully.
  exit code: 3221225477

  [0 lines of output]
  [end of output]

  note: This error originates from a subprocess, and is likely not a problem with pip.
error: subprocess-exited-with-error

pip subprocess to install build dependencies did not run successfully.
exit code: 3221225477

See above for output.

note: This error originates from a subprocess, and is likely not a problem with pip.

Press any key to continue . . .```

could I get some help?

[fossil crag]

I already tried it at least 3-4 times but 0 answer

[ruby yoke]

this discord is not part of the business office, unfortunately

[fossil crag]

ahh gotcha, thank you for your help, I thought it is

[ruby yoke]

not the business portion, which is who you need to talk to

[hallow latch]

This is what I use and get the error :/

What do you use?

I removed the Venv Folder and ran new update but still same issue :/ I watch a guy who has the same settings and work just fine :/

[karmic crown]

what error you have?

[hallow latch]

https://pastebin.com/USvUQKs5

[karmic crown]

show me his settings (screenshot)

[hallow latch]

Im testing various stuff but all the same thing :/

[karmic crown]

"I watch a guy who has the same settings and work just fine :/ "

show me what he was using

[hallow latch]

I have lowered my frames and batch size ofc

[karmic crown]

as i expected

fps - frames per secodns
number of frames - total number of frames

total number of frames = fps * duration in seconds

so if your fps is 5 and you want 3 secodnds, you need to put 15 as total number of rames

[hallow latch]

I cannot even get it to run it :/

[karmic crown]

for 30 seconds with 5 fps = 150 frames

if your settings are not good, like total number of frames = 0

it will not work

[hallow latch]

Does not matter what I do .. it will just crash with the error as I pasted :/

[karmic crown]

i only used deforum for video, but key conepts are the same, you need to specify total number of frames, fps, prompts... but also be sure to use proper image size for used model

try to close cmd, and to rerun webui first before each try

since if you get some memory related error, you might end with error even with correct settings

[hallow latch]

Everytime it happens I have to restart the CMD

[ruby yoke]

maybe consider checking how much disk space you have?

[karmic crown]

if error is realated to memory, than yes, otherwise no

[ruby yoke]

and how much vram do you have?

[karmic crown]

also valid point

[hallow latch]

I have 12 GB ( GTX 3060 with 12 GB )

I have 580 GB free out of 930 xD

It's currently running.. But with bunch of errors xD

[ruby yoke]

are your drivers up to date?

and do you have anything else at all running on that system that might be also using it?

[ruby yoke]

is the 580 on your system drive? or an extra drive?

[hallow latch]

Only Ollama and some virtual machines running linux.

[hallow latch]

It's an extra drive. I got 6 disk on my PC

[ruby yoke]

ollama uses your GPU. i would drop it for now

are you running this on linux or windows?

[hallow latch]

It's not using Ollama atm, like I am not asking it anything and the GPU is low in use

[hallow latch]

It's running on a Windows Server

This what came out of my prompt..

Kinda cool but haha 😄

[ruby yoke]

okay, my recommendation is to get rid of auto1111, install swarmUI, let it deal with all the techical stuff, and run ComfyUI from inside it. it'll make your life a lot eaiser

[hallow latch]

I like auto 1111 and been using it for like a year now :/

[ornate elk]

you have to enable this setting:

thats needed for animatediff

[ruby yoke]

any idea on that long list of errors he posted?

[hallow latch]

Do I need to remove it when not using animatediff?

[ornate elk]

you dont need to but i revert it back after im done

[ornate elk]

the onedrive folder in the paths could be a hint to the issue

onedrive synced folders causes issues with the webuis

[karmic crown]

cloud sync = sure way to hell

[hallow latch]

I could place those directly to the disc if so

I have it so that I can access the files from other devices easily 😛

[karmic crown]

you can opet webui from any devices on the same network if you use --listen

[hallow latch]

Yeah I run that but to see the images etc I have the onedrive sync so that I can see on my phone and so 🙂

[karmic crown]

than use this

and output images wherever you want

but do not put whole webui to one drive

[ornate elk]

yea better only sync the output folder

[karmic crown]

this should work, but i cant try, i do not have windows

[hallow latch]

With this .. I managed to get

I turned off Ollama and Docker

Not a single error :/

[ornate elk]

nice

[iron lantern]

FUK this is epic

hope one day i can do this

[hallow latch]

Is that Epic? haha 😄

[iron lantern]

yeah i love it

she is super hot

even my fiancee said she has nice boobs lol

[hallow latch]

I want it to be smoother 😛

[iron lantern]

teach me how to do this

[karmic crown]

smooter - means more frames - means more time to generate

so... you decide 🙂

its like when you play a game with 20 fps and 100 fps

[hallow latch]

Okay, I am testing 12 fps with 80 frames atm.

[safe inlet]

try the v3 version, it is better

[hallow latch]

I am just happy that it is not crashing 😄

[safe inlet]

ya i saw, good deal

[hallow latch]

Is this better?

[fast owl]

im having problems installing stable diffusion. getting error about the requirement_versions.txt or something, could i get some help?

[safe inlet]

the person that usually helps seems gone for the night, but check the link to the guides pinned here

[fast owl]

@safe inlet yea i asked earlier but i didnt get a response.

[ornate elk]

Can you show the full cmd log?

[fast owl]

gimme a sec, ill try it again

i closed it so i dont ahve the log

[ornate elk]

NP also whats your GPU?

[fast owl]

i had an issue with python as well, before that when downloading from the python site, it said i was missing python310.dll

[fast owl]

4080 but laptop.

[fast owl]

the micosoft store one seems to work but idk

[ornate elk]

Better dont use the ms store Python

Install the 3.10.11 64bit version, check "add Python to path" and then delete the venv folder of the webui

[fast owl]

where can i find the venv folder?

[ornate elk]

Its inside the stable-diffusion-webui folder

[fast owl]

hm, i did delete it when i tried to fresh install again

[ornate elk]

Its needed to be deleted every time you change the Python version

[fast owl]

ok hold on im trying from a completely fresh install cause i got frustrated earlier and uninstalled everything

[ornate elk]

Check my install guides from the pinned messages

For a correct setup

[fast owl]

yep followed it and got this

[fast owl]

paste bin if thats something you prefer : https://pastebin.com/Nga4V076

[ornate elk]

Open up a cmd and run
Pip cache purge
Then delete the venv folder again and relaunch the webui-user.bat

[fast owl]

i assume cmd in that folder?

[ornate elk]

Doesn't matter in this case

[fast owl]

ok, trying it again

[fast owl]

new error code, or well new old error code, i got this earlier when i tried installing it before this attempt.

[ornate elk]

Do you have any third party antivirus Programm running?

[fast owl]

not that i know of.

[ornate elk]

Okay

[fast owl]

i have windows defender on realtime protection etc like a normal human

but thats about it.

[ornate elk]

Thats okay

The pip cache purge didnt worked it seems

Can you try again?

And then delete the venv folder before launching the bat

It shouldn't be able to use cached files when the cache should be cleared

[fast owl]

okay

just to show its done

[ruby yoke]

can you reboot first, please

just to make sure memory cache is clear

[fast owl]

sure will do that too

brb

pip cleared, pc restarted, venv deleted.

trying again now

sitting on installing requirements...

[ornate elk]

Looks good, let it load

[fast owl]

ok its further than i have been so far

[ruby yoke]

yeah. don't touch it, it looks good

[fast owl]

its working boys

thank you

[ornate elk]

Perfect, no problem

[ruby yoke]

now back stuff up

[balmy shore]

whats these 'spaces' things?

[ruby yoke]

explain?

[formal trout]

Hi, I try to set a basic inpaint workflow in Comfy for SD3.5 and it went wrong

[true sphinx]

when i open my sd this is wat is on the terminal then it wont open the window for the ui

any idea why

[ruby yoke]

how did you set up the workflow? what models did you use?

[ruby yoke]

what happens if you hit the enter key?

[true sphinx]

it worked

idk y i restarted my comp and then it worked

[balmy shore]

spaces

[fair oxide]

Im trying to use OneTrainer and after training its not letting the lora be saved how do i let it be saved safetensors_rust.SafetensorError: Error while serializing: IoError(Os { code: 5, kind: PermissionDenied, message: "Access is denied." })

[balmy shore]

check read only of destination folder maybe

[fair oxide]

https://gyazo.com/b7b02ff3b779857b08f8c5b8732cd326

[fair oxide]

it wont let me get rid of the -

reopening the propertys just keeps it there

i also tryed giving full control to the folder with onetrainer in it same issue

[ruby yoke]

how much disk space do you have? and where are you trying to save it to?

[fair oxide]

145gb to D drive

[ruby yoke]

on your system drive

[fair oxide]

like the c drive?

245gb

[ruby yoke]

and you are trying to save it to what file name?

[balmy shore]

make sure to close the stuff in the folder before trying to modify it

sometimes 'file in use' will prevent changes

[fair oxide]

o let me see if i even set a file name

[serene basin]

Hey all, can someone tell me if there's a reason why a model + prompt and settings and seed would yeild totally different results in InvokeAI than in Automatic1111? Same seed, model, CFG, steps, everything. Two totally different images between Automatic1111 and Invoke.

seems counterintuitive

[ruby yoke]

which model?

[serene basin]

PicX_Real

[fair oxide]

yeah the name is just set to lora1 in the concept

[serene basin]

but also observing it for RealVision

saw some posts, on SO from 2 years ago saying the two are different, just suprised that's still the case

the "two" being Invoke and Automatic1111

[ruby yoke]

lora1.safetensors or lora1.safetensors_rust

[fair oxide]

o you have to put.safetensors in the name?

[ruby yoke]

what base model is PicX_Real trained for?

[serene basin]

SD1.5

[balmy shore]

it means the generation is non-deterministic, i think, but im not sure what that means or what the reason is

[ruby yoke]

okay well, since i have no idea what invoke has running on the backend and behind the scenes, I'm going to guess that it's messing with the output some

[fair oxide]

i put .safetensors in the name still happening

save_file(save_state_dict, destination, self._create_safetensors_header(model, save_state_dict))
File "D:\Ai\OneTrainerLora\OneTrainer\venv\lib\site-packages\safetensors\torch.py", line 284, in save_file
serialize_file(_flatten(tensors), filename, metadata=metadata)
safetensors_rust.SafetensorError: Error while serializing: IoError(Os { code: 5, kind: PermissionDenied, message: "Access is denied." })

[fair oxide]

is this for me or the other guy?

[balmy shore]

for tommy

[ruby yoke]

the other guy. one trainer's having issues writing to the directory you want it to save to

can you open a command line and see if you can change into that directory yourself?

[fair oxide]

yeah it let me

i would just use KOHYA but i get this error in it to Traceback (most recent call last):
File "C:\Users\Jay\AppData\Local\Programs\Python\Python310\lib\runpy.py", line 196, in _run_module_as_main
return _run_code(code, main_globals, None,
File "C:\Users\Jay\AppData\Local\Programs\Python\Python310\lib\runpy.py", line 86, in run_code
exec(code, run_globals)
File "D:\Ai\Automatic1111\KohyaTraining\kohya_ss\venv\Scripts\accelerate.EXE_main.py", line 7, in <module>
sys.exit(main())
File "D:\Ai\Automatic1111\KohyaTraining\kohya_ss\venv\lib\site-packages\accelerate\commands\accelerate_cli.py", line 47, in main
args.func(args)
File "D:\Ai\Automatic1111\KohyaTraining\kohya_ss\venv\lib\site-packages\accelerate\commands\launch.py", line 995, in launch_command
args, defaults, mp_from_config_flag = _validate_launch_command(args)
File "D:\Ai\Automatic1111\KohyaTraining\kohya_ss\venv\lib\site-packages\accelerate\commands\launch.py", line 852, in _validate_launch_command
defaults = load_config_from_file(args.config_file)
File "D:\Ai\Automatic1111\KohyaTraining\kohya_ss\venv\lib\site-packages\accelerate\commands\config\config_args.py", line 72, in load_config_from_file
return config_class.from_yaml_file(yaml_file=config_file)
File "D:\Ai\Automatic1111\KohyaTraining\kohya_ss\venv\lib\site-packages\accelerate\commands\config\config_args.py", line 148, in from_yaml_file
raise ValueError(
ValueError: The config file at C:\Users\Jay/.cache\huggingface\accelerate\default_config.yaml had unknown keys (['enable_cpu_affinity']), please try upgrading your accelerate version or fix (and potentially remove) these keys from your config file.

do you kno whow to fix this error in kohya?

my computar seems to hate trainers

[ruby yoke]

if you search this channel for kohya, you'll find a lost of posts with the same error. have you read the guides pinned at the top of this channel?

[fair oxide]

i havent ill take a look

[rotund flower]

How can I get rid of this error?
Commit hash: 82a973c04367123ae98bd9abdf80d9eda9b910e2
Installing torch and torchvision
Looking in indexes: https://pypi.org/simple, https://download.pytorch.org/whl/cu121
ERROR: Could not find a version that satisfies the requirement torch==2.1.2 (from versions: 2.2.0, 2.2.0+cu121, 2.2.1, 2.2.1+cu121, 2.2.2, 2.2.2+cu121, 2.3.0, 2.3.0+cu121, 2.3.1, 2.3.1+cu121, 2.4.0, 2.4.0+cu121, 2.4.1, 2.4.1+cu121, 2.5.0, 2.5.0+cu121, 2.5.1, 2.5.1+cu121)

[notice] A new release of pip is available: 24.2 -> 24.3.1
[notice] To update, run: pip install --upgrade pip
ERROR: No matching distribution found for torch==2.1.2
Traceback (most recent call last):
File "/home/dolly/stable-diffusion-webui/launch.py", line 48, in <module>
main()
File "/home/dolly/stable-diffusion-webui/launch.py", line 39, in main
prepare_environment()
File "/home/dolly/stable-diffusion-webui/modules/launch_utils.py", line 381, in prepare_environment
run(f'"{python}" -m {torch_command}', "Installing torch and torchvision", "Couldn't install torch", live=True)
File "/home/dolly/stable-diffusion-webui/modules/launch_utils.py", line 116, in run
raise RuntimeError("\n".join(error_bits))
RuntimeError: Couldn't install torch.
Command: "/home/dolly/venvs/stable-diffusion/bin/python3" -m pip install torch==2.1.2 torchvision==0.16.2 --extra-index-url https://download.pytorch.org/whl/cu121

[ruby yoke]

ignore the pip error. the issue is torch not installing

[rotund flower]

So, what do I do? How do I get it to install?

[ruby yoke]

why are you trying to install it?

[rotund flower]

Because stable diffusion wont start up without it

[ruby yoke]

what interface are you using to run stable diffusion in?

[fair oxide]

im not finding anything on my spcific issue

[rotund flower]

Mmmm? I dont know what you mean. Im using the command line 🤔

[ruby yoke]

i'm going to suggest to both of you that you switch to SwarmUi https://github.com/mcmonkeyprojects/SwarmUI/releases/tag/0.9.3-Beta install it, let IT deal with all the technical stuff, and then run comfyUI inside it and use that to geneate with stable

it will make your life a whole lot eaiser

[fair oxide]

i tryed using comfy ui as well to train a lora i was still getting an errror

[ruby yoke]

here's the main repo location https://github.com/mcmonkeyprojects/SwarmUI

[fair oxide]

swarm has a built in lora trainer or somthing?

[ruby yoke]

and instead of trying to train at home, use luca taco's trainer on replicate

[fair oxide]

u mean just use a website to train a lora?

my pc can handle it

[ruby yoke]

have you asked on the onetrainer github?

[fair oxide]

im new to github and i know people rage at you for posting the question in the wrong spot so not sure where to ask the quesiton

[rotund flower]

Can this use image as a reference?

[ruby yoke]

https://github.com/Nerogar/OneTrainer/discussions

[rotund flower]

Like upload an image and have it be modified or used as a reference?

[fair oxide]

thank you i shall try

[ruby yoke]

you can do that in comfyUI if you load a workflow with that set up or yo ucreate a workflow to do that

[rotund flower]

So... use the swarm ui to install comfyui? Is that what youre suggesting? Excuse me if these are stupid questions.

[fair oxide]

i got it nvm

[ruby yoke]

install swarmUI. let it take care of all the technical stuff. it does more than just run comfyUI inside it but comfy is fast becoming the standard for running stable diffusion and it'll run inside swarm quite happily.

we even have a #🐝｜swarm-ui channel and a #🧣｜comfy-ui channel

[fair oxide]

question is swarm just forgeui and comfy combined or somthing?

[rotund flower]

So it automatically runs comfy if I select it during installation?

[ruby yoke]

no

[crude ridge]

Hey guys hope you can help me with this

Started using comfyui, everything was working perfectly

But now when I use the "Lustify" model, my pictures end up like this

It's meant to be like this

[ruby yoke]

looks like you have the wrong VAE

[crude ridge]

How would I find out the correct Vae to use?

Im using XL by the way

[rotund flower]

@ruby yoke What do I do if the backend never loads?

[deft hemlock]

Anybody knows why pony xl slows down my pc when it reaches 95%? I run A1111

[ornate elk]

Which txt2img settings?

[ornate elk]

You have a wrong python version probably

[ornate elk]

Okay looks normal, what's your GPU?

You can try use the fp16 VAE for sdxl and pony:
https://huggingface.co/madebyollin/sdxl-vae-fp16-fix/tree/main
Its the sdxl.vae.safetensor file.

[regal kestrel]

am trying to use supermerger but it aint workin

tried to google but didnt find something

reinstall git/py3.10/a1111 and nothing

[deft hemlock]

rtx 2050 mobile with 4gb of vram, its bad but its better than my old 512mb vram laptop

i still get a slwodown when i use the sdxl/pony version and it reaches 95%

[deft hemlock]

have u tried dadding the a1111 scripyts from the github?

[regal kestrel]

my preview also stop some times for a little bit at 95% and I think its the vae

you think I should copy again from github? how do you download the scripts?

[ornate elk]

Ah okay with 4gb vram there isnt much you can do I guess

Maybe the tiled VAE extension could help

[deft hemlock]

i got it installed

but no idea how to use it

i laready got enough trauma from my old rig

u probably saw how changing the resolution could brick someones pc

[regal kestrel]

well I think you should choose no vae. or if it needs a vae make a merge with it

[deft hemlock]

it runs with both

still hits the entire pc with a slwodown

only at 95% thats the thing

[regal kestrel]

then I believe it the ram (not vram). no other idea.

[deft hemlock]

pobably

this is why i hate windows 11

its hungry as hell

like boy

12gb is almost unusable

and i cant upgrade cuz im still under waranty

[regal kestrel]

maybe if you removed some active progs?

you can also check memory use in resource monitor

[ornate elk]

If you have only 8-16gb ram you need to increase the Windows Pagefile

That can speed up the sdxl/pony gen time

[deft hemlock]

i tried but windows 11 will become a real physical entity and spank me 90s grandma style

[deft hemlock]

hmmm

how do i do that?

[ornate elk]

Here:
https://www.tomshardware.com/news/how-to-manage-virtual-memory-pagefile-windows-10,36929.html

Enable it only for C drive and set it to 16000min and 24000 max

[cloud rivet]

i need help setting up cmd

[ornate elk]

Cmd?

[crude ridge]

Any help with this?

[cloud rivet]

Yeah I was trying to use cmd to set up stable, someone on tech support was asking for my digital wallet to validate something

[ornate elk]

okay, dont give anyone information about digital wallets.
You can setup stable diffusion for free if you have a good PC (GPU with 4+gb vram)
Checkout the install guide in the pinned messages

[cloud rivet]

I got it already

Thank you

[ornate elk]

ah alright 👍

[iron lantern]

After that in the webui go into Settings - Uer Interface - Quicksettings
and there add sd_vae
then hit apply and reload ui

am in the right place ?

[ornate elk]

Yes

[clear plover]

Hello Everybody, I hope im in the right place. I am Running Linux Mint with AMD Ryzen 5 3600 and an AMD GPU 🤮

The software I am using to instll my webui is stability matrix to install Stable Diffusion WebUI Forge with DirectML (ROCM is trash apparently).

After the install I get the following error, it is unclear what my steps are to resolve

[maiden zephyr]

I just ran into an issue ive never seen before

SD crashes instantly... with no error message

(a1111)

in event viewer:

Windows cannot access the file C:0\Stable-diffusion\stable-diffusion-webui\venv\Scripts\python.exe for one of the following reasons: there is a problem with the network connection, the disk that the file is stored on, or the storage drivers installed on this computer; or the disk is missing. Windows closed the program Python because of this error.

idk what "C:0\Stable-diffusion" is

my install is on "G:\Stable-diffusion"

literally just crashed in the middle of generating after working for hours and now wont even start

[ruby yoke]

reboot your machine, and then check to make sure you're not having drive problems

[maiden zephyr]

my entire drive failed

i couldnt access anything on the drive

[ruby yoke]

ouch. turn everythign off, drive included. see if it'll start up after a full minute of bien goff

[maiden zephyr]

rebooted... and WD dashboard says the drive is 100% healthy

[ruby yoke]

so possibly there are cable issues

[maiden zephyr]

its an SSD

nvme

[ruby yoke]

there are still ports that connect. any of them dirty?

[maiden zephyr]

i can try reseating the drive

that was so weird tho

why is it working now?

im gonna back shit up before i do anything cause its seems to be working currently

[ruby yoke]

there are a lot of things that could have caused it to vanish, including memory corruption and your system just lost the pointer to it

[snow swallow]

Hello everyone. I'm trying to rum ForgeUi. I downloaded the oneclick install but I got a error related to CUDA. I'm a newbie, not sure what cuda is or how it works or if it relates to the fact I'm using a ryzen 5 with no gpu

Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Version: f2.0.1v1.10.1-previous-605-g05b01da0
Commit hash: 05b01da01f76c97011358a415820360546d284fb
Traceback (most recent call last):
File "E:\Ai\webui_forge_cu121_torch231\webui\launch.py", line 54, in <module>
main()
File "E:\Ai\webui_forge_cu121_torch231\webui\launch.py", line 42, in main
prepare_environment()
File "E:\Ai\webui_forge_cu121_torch231\webui\modules\launch_utils.py", line 436, in prepare_environment
raise RuntimeError(
RuntimeError: Your device does not support the current version of Torch/CUDA! Consider download another version:
https://github.com/lllyasviel/stable-diffusion-webui-forge/releases/tag/latest
Pressione qualquer tecla para continuar. . .

[ornate elk]

If your on Linux dont use stability matrix

You need to install the native webui with rocm drivers

[crude ridge]

Why is no one responding?

[crude ridge]

How do I find the right vae

[ornate elk]

Thats the right VAE:
https://huggingface.co/madebyollin/sdxl-vae-fp16-fix/blob/main/sdxl.vae.safetensors

[crude ridge]

Thanks man

Still not working but im going to do a reinstall

[ornate elk]

Okay np

[true sphinx]

im having it happen again and pressing enter does nothin

[ruby yoke]

yesterday, how long did you wait before you rebooted?

[true sphinx]

not that long like i made the post then restarted

then saw ur response

[ornate elk]

Delete the replacer extension and then the venv folder.
Then relaunch the webui-user.bat

[true sphinx]

ok where is the venv folder

nevermind

[paper crag]

anybody got this error recently?

[true sphinx]

ok i did that now its updating and collecting

various versions

[ornate elk]

Looks good

[ornate elk]

Nope, but your webui is outdated

[willow rapids]

hey CS1o!

whats up mate

im gonna download forge Stable Diffusion

any tips?

[ornate elk]

Hey

For nvidia or amd GPU?

[willow rapids]

lemme sned my specs

[willow rapids]

Intel Core i7 10th gen
with 1650 GTX (4gb vram)
and 16 Gb 1600mhz DDR4 ram

[ornate elk]

Ah okay, then checkout my install guide for forge