🤝｜tech-support

[loud pond]

u using SD1.5, pony, SXDL?

[fair oxide]

pony

i just want text saying Hey!

[loud pond]

oh ok, there should be some text lora's on civitai just find one you like and download it

[fair oxide]

i don`t see anything on civit

[loud pond]

lemme try and find one

[loud pond]

ok so it seems like people have been using the 'holding sign lora' then just pasting text over the sign

[fair oxide]

it works with diffrent art styles

?

[loud pond]

Not sure i havent personally tried it sorry , https://civitai.com/models/589514/folding-signxlpony

[fair oxide]

oh nah not like this :/

i have text speach cloud

[loud pond]

oh mb

[loud pond]

https://civitai.com/models/610948/comic-manga-ponyxl

what about this?

you could generate the character with the clouds and then just write in the text?

[fair oxide]

i got it alredy

i just need text

[loud pond]

https://www.youtube.com/watch?v=wB9wANpnOL0

something like this?

[fair oxide]

nvm i am just gonna copy something from internet

[loud pond]

aight sorry i couldnt help

[fair oxide]

kk

[fair oxide]

are you making nsfw things in ai by any chance ?

[loud pond]

not really if im being honest, I mainly use it for desigining game characters for my video game

[fair oxide]

do you know anything about this ?

[loud pond]

yeah ive done quite a bit of research into it, took me a while by ive managed to get character sheets, walking models, and quite a bit more

[fair oxide]

i mean about nsfw

[loud pond]

not much tbh

[fair oxide]

dm

[digital frost]

Hi, my stable diffusion is stuck at the "commit hash", computer fans are not loud whad should i do?

[loud pond]

Just so someone can solve your problem better could you send an image of your issue?

[digital frost]

and it does not go further

[loud pond]

whats your gpu?

[digital frost]

Geforce RTX 4050

[loud pond]

whats your operating system?

[digital frost]

ubuntu

on wsl

[loud pond]

oh i have no idea then, i dont use linux sorry I hope someone else is able to fix it

[digital frost]

i hope so

[ornate elk]

Why WSL?

If your on windows, use the windows Version of stable-diffusion

You won't have any advantage of using WSL with nvidia

[digital frost]

i prefer ubuntu

[ornate elk]

Ok

[indigo dirge]

I have sd 1.5 and it works normally but I want to install sdxl 1.0

@ornate elk

[ornate elk]

You have a connection error here

[indigo dirge]

What is the reason? Everything works except this

[ornate elk]

Bad wifi, or if your in Asia you need to use a VPN.
Or try change your DNS

[indigo dirge]

Yes I am in Asia

I used a vpn and it didn't work

[ornate elk]

You used one for the install or now?

[indigo dirge]

I have had a VPN for a long time, and when the problem appeared and I found some people saying that I needed to run a VPN, I used it and installed the program and the same problem appeared

[ornate elk]

Can you try without a VPN then?

[indigo dirge]

I tried and same thing

I'm tired man, I did everything I could

[ornate elk]

You may need to do a correct clean install

Using my install guide from.the pinned messages

[indigo dirge]

Where can I find it?

[ornate elk]

In the pinned messages of this channel

Or here:
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides

[indigo dirge]

Same problem appeared

Since sd 1.5 works for me.. can i make it use sdxl 1.
0?

[craggy epoch]

using Forge... ran update.bat and it says I'm up to date. However, when I launch run.bat it says I need to do a git pull?

[rugged latch]

mmm, yes, Python referred to memory address 0x00000000000, the memory could not be read

xD

i laugh as my ram slowly dies

-plays scarface - push it to the limit-

[craggy epoch]

since I used the .7z to install, where is the best place to run git pull from? Do I do it at the directory above, or within the forge directory?

[rugged latch]

u talkin bout where u install stable ?

[craggy epoch]

forge, yes... I just did a git pull from within the webui folder and it says already up to date

[loud pond]

@ornate elk sorry to disturb you but do you know how I would fix this?

[craggy epoch]

basically, if I do run.bat, I get this:

[rugged latch]

maybe ur up to date :o.

[ornate elk]

Have you tried to use an sdxl model?

[ornate elk]

To less info

Txt2img settings, cmd log etc

[ornate elk]

Have you tried the update.bat ?

[indigo dirge]

Yes I tried
But when I choose the sdxl model
The model returns to sd 1.5

[craggy epoch]

yes, I run update.bat before I start it each time

that's why I thought it was weird

[ornate elk]

How much RAM do you have?

[indigo dirge]

16

[ornate elk]

Then you need to increase the Windows Pagefile

[ornate elk]

Yea thats strange. Have you tried using git pull in the main forge folder ?

[indigo dirge]

I don't understand what to do?

[ornate elk]

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

Make sure its only enabled for C drive.
Then set it to 16000Min and 24000Max

Then restart the PC

[craggy epoch]

yep. also says up to date

[ornate elk]

Very strange.
Maybe you need to redownload it

[indigo dirge]

thus?

I have windows 11

Press ok?

[loud pond]

alright

[indigo dirge]

@ornate elk It didn't work.

[rugged latch]

lil random tip for those who may experience it, if u ever run into a runtime error where stale could not allocate enough video memory for a tensor or sum, save the prompt and restart pc

make sure to click restart as that will flush ram, once done, the prompt will work

[loud pond]

yeah ive noticed that

[loud pond]

why does that work btw?

because when ever i run a big prompt it wont work after an hr of using my pc but when restarted it will work

[craggy epoch]

on the 3rd attempt using the .7z... might have to do a manual git clone if it doesn't work out

[ornate elk]

Can you share the full cmd log?

[neon orchid]

Guys, a question.
How can I tell if a Civitai file is for Embeddings, for Vae, for Lora or for Stable-Diffusion?

[loud pond]

should tell u in the description

[neon orchid]

They recommended me 2 Civitai files to improve the drawings generated by Stable Difussion, but on their pages they don't show which folder they are assigned to.

[neon orchid]

No, he doesn't tell me 😦

[neon orchid]

I don't know in which folder this is designated

[loud pond]

checkpoint merge is stable diffusion

(thats how its worked for me anyways)

if you put any checkpoint-merge file into your stable-diffusion folder you should be ready to go

[loud pond]

did that help?

[neon orchid]

I'm looking for that folder

[loud pond]

\models\Stable-diffusion

[neon orchid]

ohhhhh

i don't read the message

sorry

[loud pond]

all good, is everything sorted out now?

[neon orchid]

yes, i think

I'll try if it works.
Thanks 😄

🫂

[loud pond]

alright message again if you have any more issues!

[rugged latch]

bc boop

[sonic arrow]

Does anyone know how Copilot works?

[ruby yoke]

yes? a bit more specific question, please

[solemn moon]

Same settings, driver updates, now SD freezes more often than not o0
Can the Nvidea bros get on it TIA

[sly pike]

there is numerous numbers of files in my 'lora' folder, but only half of them show in sd when i'm generating images. how can it be fixed?

[craggy epoch]

with these specs, how can I tune my system for SD Forge, A111, or others? I'm finding that I get like 4 - 5 it/s while I see people with similar configurations getting 9+ it/s on similar or the same models.

[glad fiber]

i get around 3s/it on a1111 and less than 1s/it on comfyui

gtx 1650

so i belive it has something to do with specif uis

[craggy epoch]

I've been playing with Flux, which A111 doesn't support yet. However, even using the realcartoon checkpoints in A111, I still seem to get about 1/2 the it/s that others post

[sonic arrow]

do you know how to access like old Microsoft designer or at least access like a saved album on there

[ruby yoke]

unfortunately, no. microsoft probably doesn't have anything from the old version accessable to anyone. you can try contacting them

[sonic arrow]

ok

[ornate elk]

What's in your webui-user.bat?

[strange quest]

im having a problem with this cuda error

i'm not sure i understand the answers of disabling cuda malloc that im reading

does anyone know what i need to add to the .bat?

@ornate elk SEC discord suggested i ask you how to fix this

[ornate elk]

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

[ornate elk]

SEC discord ?

[vocal burrow]

cuda malloc thingy doesn't really matter. It's a debug parameter for developers (you should not use it).
What s important is the fact that you re running "out of memory (vram)"

[versed imp]

Anyone can tell me what is about this error please?

[ornate elk]

Can you show the full cmd log?

[versed imp]

Yes,is very big, should i put in notepad or something?

[ornate elk]

As txt yes

[ornate elk]

You installed the webui the wrong way (by downloading it as zip file) so you won't get any new updates.

Also you installed the roop extension which is outdated since over a year.

[versed imp]

For this reason it can cause many things to stop working, like prompts, Lora and things like that?

Not working as it should

[ornate elk]

Yes

Better is to reinstall the webui the correct way with my Guide from the pinned messages

[versed imp]

Can u send to me please?

[versed imp]

there told to find StableSwarmUI and dont have any

[vocal burrow]

what do you mean ?

[versed imp]

Go into the StableSwarmUI folder and run the launch-windows.bat

[vocal burrow]

You choose whichever tutorial you prefer. Be it StableSwarmUI, auto1111, comfyui, etc

[versed imp]

Dont have this folder

Oh

[vocal burrow]

according to your previous log, you seem to be using auto1111 not stableswarm

[versed imp]

for me is web ui

So from the other instalation, which folder i can copy for the new install?

To not lose all configs

[ornate elk]

The webui is Automatic1111

Follow that guide

[versed imp]

Oh sorry i mean bc have Lora, check point and anothers things

[ornate elk]

You can move them over

[versed imp]

Can i move all this folder to the new install?:

[ornate elk]

Only move the stable-diffusion and lora folder over

And the vae

The rest is empty or will be recreated

[versed imp]

venv "C:\AI\stable-diffusion-webui\venv\Scripts\Python.exe"
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Version: v1.10.1
Commit hash: 82a973c04367123ae98bd9abdf80d9eda9b910e2
Launching Web UI with arguments:
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
[-] ADetailer initialized. version: 24.9.0, num models: 10
Loading weights [2d0d06f4b8] from C:\AI\stable-diffusion-webui\models\Stable-diffusion\aicomic_v10.safetensors
Creating model from config: C:\AI\stable-diffusion-webui\configs\v1-inference.yaml
Running on local URL: http://127.0.0.1:7860

To create a public link, set share=True in launch().
C:\AI\stable-diffusion-webui\venv\lib\site-packages\huggingface_hub\file_download.py:1142: FutureWarning: resume_download is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use force_download=True.
warnings.warn(
Startup time: 12.4s (prepare environment: 3.2s, import torch: 3.9s, import gradio: 0.8s, setup paths: 0.8s, initialize shared: 0.3s, other imports: 0.5s, load scripts: 1.7s, create ui: 0.6s, gradio launch: 0.6s).
Applying attention optimization: Doggettx... done.
Model loaded in 3.7s (load weights from disk: 0.9s, create model: 0.5s, apply weights to model: 1.9s, calculate empty prompt: 0.2s).

everything is good now right?

[vocal burrow]

seems good so far, but what s your gpu ?

[versed imp]

RTX 4060 8GB

Why?

[vocal burrow]

Then you want to add --xformers --medvram-sdxl --no-half-vae command line args in your webui-user.bat ( like probably stated in the guide )

[idle plank]

Hi I've got some problem installing stable diffusion 1.5, I don't find any help that get me to the final install, I maybe have some problem with Pytorch but I think I already resolve this and it's still doesn't work :/ Someone can sharescreen with me during a call please ?

[vocal burrow]

cant sharescreen but what do you mean by "installing stable diffusion 1.5" ? What are you trying to install exactly ?

[idle plank]

I'm trying to install the gitHub Automatic 11111

I downgrade to a lesser version of pytorch and it works but I can't launch the software still

It actualy says that it can't found Python whereas I already installed it on the same sdd

[vocal burrow]

what hardware and OS ?

[ornate elk]

Can you share the cmd log?

[rugged latch]

bOoP

open webui-user.bat

in notepad

and where it says "python:"" "

add the directory

should look like dat, replace mxz with ur user name

and "C:" with wut ever ur drive letter is

if python its self is being dumb, u may need to add environment variables

but, shouldn't have to

lemme know if it worked =3

[vocal burrow]

To me, the fact that they re talking about "downgrading pytorch" indicate that they re either not on windows or totally lost. Let s make sure we know what we re working with first.

[rugged latch]

i dunno, to me it looks like their skipping the step to tell stable where python is, as they said "it actually says cant find python" and that they installed it

meaning, stable is launching, but cant find it

buh oki, they may be on linux, who knows xd, in that case, wut i sid wouldnt work anyway bc linux handles drives diffriently

[thorny iron]

What does this mean

Oh, 'no half vae' was missing in my bat.. whoops

Never mind, still doing it

[vocal burrow]

what s your gpu ?

[thorny iron]

4080 laptop 12gb

Maybe my picture is too big?

I didn't have problems before.

[ornate elk]

which settings, did you used?

[thorny iron]

I'm inpainting, and some reason the error happens when I change to a different model.

[thorny iron]

Which settings do you mean just to be sure

[ornate elk]

make sure you set inpaint area to "mask only"

[signal sigil]

Hey guys!
Just wanted to ask this question. When training a LoRa in two separate sessions.
The folder that the first session of training made. Do i need to put the LoRa model from the first training session inside that folder for it to continue training?
(im using the LoRa Easy Training Script btw.)

[thorny iron]

Ah those settings, I put a screenshot

The error seems to happen if I change models though..

I get a black square, then the error comes

[vocal burrow]

what model are you using ?

[thorny iron]

I closed everything for the night sadly, really tired. I'll try again later! 💙

[signal sigil]

Hey m8. Might you know the answer to my conundrum at all?
"Just wanted to ask this question. When training a LoRa in two separate sessions.
The folder that the first session of training made. Do i need to put the LoRa model from the first training session inside that folder for it to continue training?
(im using the LoRa Easy Training Script btw.)"

This issue has plauged me the last few times i have tried to continue training a checkpoint. It doesn't seem to be leaving any indications that it is a continued trained checkpoint after it has finished. Almost like it has trained from scratch. So i just wanted to rule out if my qustion was a possibility?

[loud pond]

@ornate elk pls help

if you need anything else pls inform me

[gleaming prism]

hello idk why but i get this error today yesterday was working fine

[ornate elk]

Can you show the full cmd log?

[gleaming prism]

and i get only a gray image even just using the model no loras i am using FlUX 1 dev

[ornate elk]

Flux doesn't work in Auto1111 currently

[gleaming prism]

weird because yesterday when i used it it worked

[ornate elk]

It can't work maybe it switched back to the old model

[loud pond]

also @ornate elk how did you learn to fix all of this?

[ornate elk]

That could have been done in forge as its based on Auto1111

Auto1111 has no support for flux at all since the last update is from before flux release

Only Forge webui, SDnext and Comfyui can use it

[ornate elk]

Set the hires steps to 10 ans the denois to 0.5 and try again

[loud pond]

alright

[ornate elk]

What's your GPU again?

[loud pond]

Amd radeon 480

[ornate elk]

Then you need to enable the tiled VAE extension too

If you want to upscale

[loud pond]

how do i get that?

[ornate elk]

In the extensions tab

Under available, load from

Then search for tiled

[slow dove]

why im getting this error?

[ornate elk]

Mostly that error appears when using a lora or embedded that is not compatible with the checkpoint

[slow dove]

yeah, it's LoRa

[loud pond]

what should i enable? @ornate elk

i got the tiled VAE extension

[slow dove]

how and where i camn check what checkpoint i need? i got couple of models from civitai

[ornate elk]

Only enable tiled vae

[ornate elk]

You have to check the loras info box at civitai

There you see what its made for

[slow dove]

like base model?

[loud pond]

What would you suggest?

@ornate elk same error

[ornate elk]

Yes

[ornate elk]

Which model you tried?

[loud pond]

blendermix

[ornate elk]

Try set upscale by 1.2

[loud pond]

alright doing it rn ill tell you how it goes

[loud pond]

@ornate elk same error

[ornate elk]

Can you try an other upscaler?

Or without an upscaler?

[loud pond]

ok ill do both

[loud pond]

used a different upscaler but didnt work ill try no upscaler now

[loud pond]

@ornate elk so it semi worked without an upscaler

An image was produced but there is a massive Grey box on the image

[ornate elk]

Oh then disable the tiled vae

[loud pond]

ok ill attempt it again without the tiled vae

[ornate elk]

Or try to set the tile encoder to 768 and decoder to 128

[loud pond]

ok ill do both!

[loud pond]

disabled the tiled vae and it worked

[loud pond]

gonna try with this now

[loud pond]

both ways worked

thanks! so was the type of upscaler i used the issue? @ornate elk

[ornate elk]

Can you try the Resrgan 4x+ Upscaler?

That should work normaly

[loud pond]

Oh ok sure

[gleaming prism]

I am not joking it worked yesterday with automatic1111

[loud pond]

@ornate elk stopped working with Resrgan 4x+

[iron ice]

I just installed stable-diffusion-webui and been playing with the models. However, I really like this one from Replicate - https://replicate.com/adirik/t2i-adapter-sdxl-depth-midas is possible to use that locally in the webui? Or do they keep those models private so they can charge?

[ornate elk]

You can use this model after you installed the Controlnet extension for the webui

[iron ice]

I got it, that's the extension I've been playing with

[loud pond]

@ornate elk does that mean its the upscaler which is broken?

[iron ice]

It's a depth model, right, is it possible to download it ?

[ornate elk]

yes

[iron ice]

I got ControlNet's default one control_depth-fp16.safetensors

how do I download one from the replicate link?

[ornate elk]

maybe its one of these:
https://civitai.com/models/136070?modelVersionId=154942

ah nope, here it is:
https://huggingface.co/lllyasviel/sd_control_collection/tree/main

[iron ice]

t2i-adapter_diffusers_xl_depth_midas.safetensors

nice one.

How did you find it ?

[ornate elk]

i have a lot of saved links to everything of stable diffusion ^^

[iron ice]

Hmm,. I'm not getting good results compared to the default one from ControleNets control_depth-fp16.safetensors

Totally different to what I've gotten on replicate

what's the difference?

[ornate elk]

could be anything

we dont know what replicate does in the background

or you have to play around with the controlnet settings more

[iron ice]

k fair.

It doesn't seem to really respect the depth mask

like completely ignores the perspective

[loud pond]

@ornate elk could you please explain how this works? I would be very greatful if you could thanks in advance!

[rugged latch]

whaaaaaaa

a laptop with a 4080

;o

kinda feckin laptop u got xD

that bish must be a feckin lapwarmer fr

that or sound like an A320 bouta take off

[iron ice]

I'm trying to use sd3 VAE model but getting this error when trying to select it

[rugged latch]

some newbies join server and come here be like "hello, whats stable diffusion"

xd

[iron ice]

under 23? 🙂

[solemn atlas]

Hii, how do I do that?

[ornate elk]

Launch windows on the 4tb sdd

Then format the other drive (1tb)

[solemn atlas]

Okay, It's the iso right? I downloaded it on my pc (not usb) is that right?

[ornate elk]

No I mean you boot into windows

With the 4tb

[solemn atlas]

yes, but it's this right?

I have my 4tb plugged in and downloaded the iso

[ornate elk]

Nope. I dont have time rn sry.

Have you meanwhile tried asking someone you know if they can help you? Because its easier to fix this in person than remote

[loud pond]

@ornate elk

found the solution to the problem I was having

https://www.reddit.com/r/StableDiffusion/comments/1ef6w5b/error_when_using_upscaler_on_amd_gpu/

Apparently its an issue the the AMD GPU, there is some talk about having to downgrade something but idk rn

ill do more research into it i just wanted to let you know lol

thanks a lot for your help!

[loud pond]

tbh i think i should just buy nativida to save me the trouble xd

[pale magnet]

hey there! i just entered the server. i'm in look of help to finetune a model to a small dataset i've created for college. i never used stable diffusion before, and i'm getting totally beat by the errors lol

im trying to use a LoRA (?) but with DiffusionPipeline, but i get this:

ValueError: Pipeline <class 'diffusers.pipelines.stable_diffusion.pipeline_stable_diffusion.StableDiffusionPipeline'> expected {'scheduler', 'text_encoder', 'unet', 'feature_extractor', 'safety_checker', 'tokenizer', 'vae'}, but only {'scheduler', 'text_encoder', 'feature_extractor', 'safety_checker', 'tokenizer', 'vae'} were passed.

i didn't find any notebook that is just like "insert your folder here" so i'm trying to make one jupyter notebook by myself, but it's not my area of expertise

[gentle dew]

hello, im trying to download orangemix as a beginner is there a common guide i can follow to streamline the process? thank you in advanced

[ornate elk]

You can download the model and put it into the models/stable-diffusion folder

[gentle dew]

whats the best webui?

[wicked edge]

Anyone knows what is this error, please help me.
*** Error loading script: main.py
Traceback (most recent call last):
File "D:\Stable-diffusion\stable-diffusion-webui\modules\scripts.py", line 515, in load_scripts
script_module = script_loading.load_module(scriptfile.path)
File "D:\Stable-diffusion\stable-diffusion-webui\modules\script_loading.py", line 13, in load_module
module_spec.loader.exec_module(module)
File "<frozen importlib._bootstrap_external>", line 883, in exec_module
File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed
File "D:\Stable-diffusion\stable-diffusion-webui\extensions\openpose-editor\scripts\main.py", line 14, in <module>
from basicsr.utils.download_util import load_file_from_url
ModuleNotFoundError: No module named 'basicsr'

[modest tapir]

I have been told there's some kinda steps I need to take to get an AMD CPU to play well with Stable Diffusion Forge/Flux etc rendering, I have an RTX 4090 - is there a link or guide someone can point me to to get that done? I'm not sure what the exact issue is as my render times don't seem that far off what others have reported with similar hardware. Only oddity seems to be my system RAM usage - I upgraded to a 2x48GB kit because the 2x16GB kit I had was getting fully-utilized by Python during loading stuff before rendering started & the system would lag for about 1-2 minutes before the first render actually started (and loaded the VRAM on the GPU). That issue doesn't exist now that I have plenty of RAM, but apparently my system RAM usage is a bit high compared to what it should be...? I don't know. lol I'm still learning this stuff.

[ornate elk]

Thats an error from the openpose editor extension

go into the stable-diffusion-webui folder.
There click into the explorer bar (not searchbar)
type cmd and hit enter.
then you have to run these two commands:
venv\Scripts\activate.bat
pip install basicsr
then relaunch the webui-user.bat

[ornate elk]

Automatic1111

[ornate elk]

No you have nothing to do as stable diffusion doesn't care about the CPU brand

Additional steps are only needed for AMD GPUs

What's in your webui-user.bat?

[modest tapir]

Nothing special I don't think... let me post the code if it'll let me...
"
@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=

@REM Uncomment following code to reference an existing A1111 checkout.
@REM set A1111_HOME=Your A1111 checkout dir
@REM
@REM set VENV_DIR=%A1111_HOME%/venv
@REM set COMMANDLINE_ARGS=%COMMANDLINE_ARGS% ^
@REM --ckpt-dir %A1111_HOME%/models/Stable-diffusion ^
@REM --hypernetwork-dir %A1111_HOME%/models/hypernetworks ^
@REM --embeddings-dir %A1111_HOME%/embeddings ^
@REM --lora-dir %A1111_HOME%/models/Lora

call webui.bat
"
That without the quotes, should I change anything?

Also I'm asking because I'm seeing pretty cray cray system RAM usage during loading/unloading of data before it's loaded on to the GPU VRAM... like I've seen anywhere from 20-70GB of System RAM usage by Python/SD depending on how many LoRAs & such I'm using at once, is that normal? lol

[ornate elk]

Thats not normal

What model did you used?

The bat looks normal. And your using forge webui instead of auto1111

[modest tapir]

Yes, I'm using Forge, and the system RAM usage varies on the model/checkpoint, on the full Flux Dev that's Fp16 is when I see the most usage though, especially if I'm using bigger LoRAs

I assume that these settings are all "normal" - I change the GPU weight as needed:

You can see the CPU & drive usage here too (should be noted, T: drive is where Forge is installed) - that's a fresh load of the WebUI then going in to a render with flux_dev loaded & a single LoRA that's on the beefy side (several hundred MB not anything too crazy) - that's why I'm confused cuz I've seen a few things about usage well above 32GB being abnormal and I don't know what I'm doing wrong/what's set wrong lol

[ornate elk]

You need to use the smaller flux models

Dont use the 23gb one

[ornate elk]

The GPU weight needs to be larger than the model

If not it will use RAM

So best for a 4090 would be GPU weight of 22gb

And then using the fp8 flux 16gb model

[half island]

Hoi, for you who have made loras with kohya, how many steps do you guys set it to be per image? As i set mine to 1300 steps 1 epoch for 9 images, tested with 128 netowrk dim, 64 alpha, also tested 32 netowrk with 16 alpha for styles with adamW , but neither gets a good result at all :S

And if you need number_model-name for source images, should that be steps per image, or overall steps?

Need above 2.3 model strength in comfyui to get it to apply anything, and by theh, image's already fried.

[elder mauve]

getting error with AIO aux preprocessor and control net in comfyui

does this work for anyone?

[mystic basin]

help pls

[rugged latch]

ari

im extremely confused

why did u download an iso thats jus labled "windows"

when u was talking about formatting an ssd to lose the UEFI partition

<v>

[ornate elk]

Can you share the full cmd log?

[mystic basin]

i can

wait now it's working

i tried something, it didn't work so i tried the same thing again to get the log and now it work

wait nvm

[rugged latch]

is so annoying when this happens

[mystic basin]

it did not work

[rugged latch]

sum parts in there say the repository could not be found

maybe link changed or is gone

[mystic basin]

probably, i might have to switch to confy or swarm

[vocal burrow]

what s your gpu ?

[mystic basin]

amd radeon rx 6850mxt

could it be because my gpu is too weak? it's a laptop gpu (my laptop is plugged in btw)

[vocal burrow]

it should work, but I recommend using the zluda install instead of directml. Check pinned guide.

[mystic basin]

alright, i'll try zluda, thanks for helping me

[rugged latch]

can also try disable tile upscale

error happens during that process

my stable dunt like tile upscale either

[mystic basin]

this?

[rugged latch]

my oictures come out black xd

gib a sec, lemme start mine

[mystic basin]

this is causing the error i think, it work when i don't use it

[rugged latch]

Ya

[vocal burrow]

still recommend going zluda, much better performances.

[rugged latch]

stable, casually taking its time to start xd

this is why u diont install stable on a 50mb/s hdd

[mystic basin]

i'll try it later, i don't have the time right now

but thanks for informing me

[rugged latch]

dis wut i was talkin bout

when thats on for me, it causes my pics to be jus, a black void, or sum times they look like, demons from the hell demission

[mystic basin]

idk where to find that

[rugged latch]

but if not, go to settings tab, at the top type "tile" and youl see it

[vocal burrow]

that s in the settings tab, if you don t know where it is, it is very unlikely that you ve changed it.

[rugged latch]

ya

and if u said, not using hires fix makes ur pics oki

then it must be dat

[mystic basin]

i never changed anything in setting

could this be the problem? it's supposed to upscale the image but i can't even make an image when the upscaler is on, i click on generate then it give me an error message

i don't know if it's that, someone told me to use R-ESRGAN 4x+ once

[rugged latch]

have u tried

make the image first

then

img2img upscale

[mystic basin]

nope

let me try

[rugged latch]

try dat xd

thats how my bestie does it

[mystic basin]

it didn't

i think it's just the upscaling thing i chose

i'm trying with another one and it seem to work so far

[ornate elk]

Let me know if you need help with the ZLUDA setup

Its the best we have for AMD on windows right now

[ornate elk]

Directml has an issue with Upscaling currently.
Set the upscaler to None should work

And hires steps to 10

[mystic basin]

oh that's why it didn't work

i'll switch to ZLUDA when i can, in a few hours

thanks for fixing my problem

and have a nice day

[ornate elk]

No problem, you too

[rugged latch]

gud thing im still on 1.7.0

:3 xD

[median geode]

hello guys, anyone know what extension is needed if i want to Generate prompts with pictures

in webui

i heard that tagger can no longer be used

[elder mauve]

@ornate elk do u use comfyui?

[ornate elk]

only for tests with flux

[elder mauve]

can u test if this works for u

[hazy plank]

I can't access Stable Assistant, after Google login I keep seeing a full black browser screen. Clearing cache also doesn't work. Anyone knows how to fix this?

[sterile coyote]

Dunno if this belongs here, but can I change an existing image into cartoon style without needing to describe the existing image in the prompt, just sth like "turn into cartoon"?

[clear silo]

i need help i find a method for use automatic1111 on a colab with kaggle, but when i use the local link, the local rejects the connection

[rugged latch]

wut did u change, im confused

[elder mauve]

getting "Cuda out of memory , tried to allocate 28GB " while trying to upscale , i have 16GB vram , what to do?
this is the workflow

[mystic basin]

i was told to switch from DirectML to ZLUDA

[loud pond]

Could someone help pls. Whenever I generate an image everything comes out perfect except for the face and (sometimes the body).

[wet flame]

'list' object has no attribute 'replace'

[balmy monolith]

Some of the models have the External Generator tag for the prompt to generate an image, what is this please?

[wet flame]

found my issue

hmm still same

works when i remove show any node

[craggy epoch]

apologies, got busy yesterday and today, but here is my webui-user.bat for forge:

`@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= ^

--update-check ^
--xformers ^
--cuda-malloc ^
--ckpt-dir "D:\image\stable-diffusion-webui\models\stable-diffusion" ^
--vae-dir "D:\image\stable-diffusion-webui\models\vae" ^
--lora-dir "D:\image\stable-diffusion-webui\models\lora" ^
--embeddings "D:\image\stable-diffusion-webui\embeddings" ^
--listen ^
--enable-insecure-extension-access

@REM Uncomment following code to reference an existing A1111 checkout.
@REM set A1111_HOME=Your A1111 checkout dir
@REM
@REM set VENV_DIR=%A1111_HOME%/venv
@REM set COMMANDLINE_ARGS=%COMMANDLINE_ARGS% ^
@REM --ckpt-dir %A1111_HOME%/models/Stable-diffusion ^
@REM --hypernetwork-dir %A1111_HOME%/models/hypernetworks ^
@REM --embeddings-dir %A1111_HOME%/embeddings ^
@REM --lora-dir %A1111_HOME%/models/Lora

call webui.bat`

one thing I noticed is that when I do run a batch (could be 1 batch of 1, or 5 batchs of 20, doesn't matter), it seems like Forge has to clear out a bunch of RAM from the GPU (4090Ti Super @ 16GB) before it will begin doing anything. Which is fine, but it iterates through multiple runs of a large chunk >1024MB, then starts doing 2 or 3 batches of just 1024MB until it's cleaned enough out to move forward.

[modest tapir]

Ah, that makes sense! I will try that here in a few and report back if the issue persists.

[terse sun]

so i'm just being curious here, i have seen sd models seems to be very vram hungry but uses very little of gpu, since i work in repair shop i have been thinking if i say took an older gpu and replaces it's vrams with higher capacity ones, would the models works better than newer gpus with less vram?

[mystic basin]

i've switch from directML to ZLUDA and i get this error, i don't understand. can anyone help me?

that might be why i have problem. i just need to find what i did wrong

i might have set one of the path in the variable wrong

[mystic basin]

i tried changing the variable path for HIP_PATH to C:\Program Files\AMD\ROCm\6.1\bin\rocblas.dll
still doesn't work

it seem like i try to find ZLUDA at the wrong place but idk what to change since i already have put the two variable needed like the guide told me

[mystic basin]

i just realized i named the folder wrong but that's not the problem

[ornate elk]

Your path settings are incorrect

Click on these two buttons then, add the 2 paths their.
Change the HIP_PATH value back to its default.

[solemn moon]

Would be cool to if there was some sort of notification that said which extensions are being faulty o0

[solemn moon]

could a code bro help me clean this up

[ornate elk]

What's your GPU?

[solemn moon]

1660ti

i deleted Auto for Forge but went back to a mess on Auto

[ornate elk]

Okay then remove --precision full and --opt-split-attention from the webui-user.bat
And change --lowvram to --medvram

And then enable FP8 mode for SDXL in the settings.
Then you can use sdxl models faster

You can also remove FaceFusion if it doesn't work

[solemn moon]

alot cleaner thanks💚

[ornate elk]

NP, also test if reactor works

[solemn moon]

doesnt seem to work

t'is a sad day gentz, ive lost my power to mess with muh frenz

[ornate elk]

Then you need to install insightface

Here is a guide:
https://github.com/Gourieff/sd-webui-reactor?tab=readme-ov-file#insightfacebuild

[solemn moon]

git clone to extentons lg!!!

[ornate elk]

Then reactor will work again

[ornate elk]

No

Its a step needed for reactor

[solemn moon]

okay trying again

so for this:

i get this:

.\venv\Scripts\activate : File C:\Users\o0\stable-diffusion-webui\venv\Scripts\Activate.ps1 cannot be loaded because
running scripts is disabled on this system. For more information, see about_Execution_Policies at
https:/go.microsoft.com/fwlink/?LinkID=135170.
At line:1 char:1

• .\venv\Scripts\activate

•   + CategoryInfo          : SecurityError: (:) [], PSSecurityException
    + FullyQualifiedErrorId : UnauthorizedAccess

just left that part out & see if it works 😅

[ornate elk]

You have to activate the activate.bat

With the cmd

[mystic basin]

Which one?

[fervent gorge]

does anyone know if the api for a111 has a max prompt length?

?

Im using dynamic prompting and i keep getting errors that say { ...

[ornate elk]

Both

[mystic basin]

i don't know what i did wrong

i think i wrote each path correctly like it's said in the guide but i'll check again

yep i don't see any difference

i need to change the path but i'm not sure what to change

[vocal burrow]

Did you install HIP SDK ?

[mystic basin]

yes

my amd folder,

if i'm not wrong, it's the ROCm folder that is related to HIP SDK

my variable should be fine i think

[vocal burrow]

if you run echo %HIP_PATH% in cmd what does it show ?

[mystic basin]

i hided the first part by the way since there's my last name

[vocal burrow]

what have you done when editing your variables at step 4 ?

%HIP_PATH% is supposed to return a directory, not %HIP_PATH%bin

so you probably overwote whatever was in there somehow

[mystic basin]

i checked what was written on the guide and copy pasted, i can try to delete the ZUDLA variable and the HIP_PATH one and recreate new one

[vocal burrow]

you re not supposed to create new variable for them

[mystic basin]

i created new variable

[vocal burrow]

you re supposed to add c:\ZLUDA and %HIP_PATH%bin to the already existing PATH variable

delete the ones you created

[mystic basin]

oh

these one? i'm not sure about the part of the already existing path variable

[vocal burrow]

no

there is a variable called PATH

(got to go for now)

[mystic basin]

the only variable called path i have is for oculus

[vocal burrow]

that one, and no it does not only hold oculus stuff

it holds a bunch of stuff, add c:\ZLUDA and %HIP_PATH%bin to it

[mystic basin]

oh i see

[vocal burrow]

see ya

gl

[mystic basin]

see ya

thanks for the help

it should work now but i got to go too

so i don't have the time to try now

[ornate elk]

You also installed the wrong hip sdk version.
Your card needs the HIP SDK 6.1

[undone wren]

Guys, does anyone work with scripts in stable? I have a problem, the script works correctly, but for some reason upscale doesn't want it(

[solemn moon]

its working, i owe you like 3 now 👊

[hallow latch]

I am having issue when downloading lora's atm .. I am running stable diffusion on another PC... I finds all other lora's and so on.. but I have even put files manually in there and still can't find it.. Could someone guide me to what I should look for ?

Can it because I have it as listen .. the command line so that I can access it from another pc?

I ran it without the " --listen" but still cannot find the Lora I downloaded :/

same with another Lora that is in the folder but won't show up.

[solemn moon]

have you tried GPT bro

just helped me

[hallow latch]

I have tried before and gpt is not helping with this issue.

I went over GPT and did not solution here and checked what gpt said

[vocal burrow]

you re not trying to load sd1.5 lora while using a sdxl model ? (or vice versa)

[mystic basin]

I did download 6.1

Unless i didn’t

[vocal burrow]

looks like you have two of them. probably ok, I don t have an AMD gpu tho so I can t vouch for it.

[mystic basin]

Well I downloaded an earlier version a few month ago but downloaded a more recent one yesterday so idk if it will work

I’m getting home in around 6 hours so I’ll be able to test if it work once I get home

[hallow latch]

Does it matter what Lora it is? The problem is that I can't see the lora's..

[vocal burrow]

auto1111 won t show incompatible loras by default

so yes if you re trying to use a sd1.5 lora while having loaded a sdxl model it is normal for it to not show up.

[hallow latch]

How do I know if it's an sd 1.5 lora or not?

[vocal burrow]

you check wherever you downloaded it from, there should be a readme/description.

[vocal burrow]

also I have no idea what that droplist is.

[hallow latch]

It's from Civit AI Browser extension

[elder mauve]

getting this error

[elder mauve]

.

@oblique meadow

[vocal burrow]

what are the dimensions of your image ?

[terse siren]

Will Flux ever get to the point where just making one fucking picture won't compleltey nuke my PC?

[ornate elk]

Depends on your GPU and RAM

[wet flame]

is this the place to ask how to keep same colour composition?

if i mention the colours it keeps it?

[mystic basin]

i have new error now that i fixed the variable problem i had

i think i know how to fix this

i had two version of HIP SKD, i deleted the oldest one and this happened, i need to redo the git clone part and i should be fine

sorry for my bad english, english isn't my first language

nvm it wasn't that

idk why the TensileLibrary.dat isn't there, i'll try to redownload hip sdk

nope, didn't fix my problem

[ruby yoke]

what is the exact issue?

[mystic basin]

it says it can't find a certain file

i have hip sdk on my pc

[ruby yoke]

open up file explorer, go to the directory it is looking for. if the directory is there, see if the file is in it

[mystic basin]

it isn't, that's the problem

that's the first thing i did

[ruby yoke]

okay so now you need that file, and you then need to put that file into that directory by hand

where does that file come from?

and how much space do you have on your C: drive?

[mystic basin]

HIP SDK, it's something for amd gpu and you need it for ZLUDA to work

i downloaded it from the amd website

not much space left but it should be enough

[ruby yoke]

okay, take a step back and a deep breath. and then go clear out enough space to have a couple hundred, not 69, free

you barely have enough space for your sytem to even run

[mystic basin]

i'll find useless thing to delete

[ruby yoke]

or move to some other drive. like all the models you have. move them - and then tell whatever interface you use where to find them

[mystic basin]

i only have one drive so deleting thing is the only option

still haven't finish to delete some of the thing

[ruby yoke]

okay. first uninstall what you just tried to install. get it all out of th eway and gone. you'll need to do it over once you have room

[mystic basin]

i should be fine now

[ruby yoke]

you should be providing the new install doesn't run you back into the red

make sure everythign is uninstalled, including the install files

[mystic basin]

i should delete all these amd hip sdk related file right

[ruby yoke]

once you have all of that cleaned up, download stuff again. no idea if it downlaoded correctly with your drive that full

[mystic basin]

i'll try that

[ruby yoke]

and take your time, don't rush. think each step through

and check your disk space every little bit as you do this

[mystic basin]

i should also delete the temporary file if there's any?

i should have enough space now

even more now

now i should redownload the installer and try again?

nope still the same error

[mystic basin]

it's probably because it's trying to use my igpu, also when it download it since hip sdk and ROCm doesn't support igpu

i don't know how to make sure it use my gpu when downloading, i mean go to my gpu and not my igpu

apparently i need to go in the bios to desactivate my igpu redownload and all but after i don't know if i can reactivate my igpu or not

or do this

idk if my gpu is 0 or 1 so i'll try both
but i'm pretty sure my gpu is 1

i'll need someone to help me because this ain't working

[ruby yoke]

okay, you'll need to wait for @ornate elk to log back in I think

[mystic basin]

or any other guides

[ruby yoke]

he wrote the guides. he's the one that is going to need to help you with this. however you could try asking on the https://community.amd.com/t5/rocm-discussions/bd-p/amd-rocm-discussions group

[mystic basin]

yes i could but i don't think this will help

i've tried a lot of thing i found related to my issue but it didn't fix it

it might be me that is doing things wrong

[ruby yoke]

i would just wait and talk to @ornate elk once he's back online if it was me

[mystic basin]

i think that's what i'll do

since i don't know how to fix my issue

[ornate elk]

That error means zluda found your CPUs igpu before your dedicated GPU.
You have to go into the device manager and there under display adapters, disable the Radeon TM Graphics.
Then relaunch

Or you can add: HIP_VISIBLE_DEVICES=1
to the webui-user.bat
Then save and relaunch.

[fair oxide]

im using comfy ui cogvideox is there a temp folder or a folder that has all the frames of the created video that cogvideox makes any where ive looked in the output and input folders and there not there or does it auto delete after generation?

[worldly trail]

Hey, I'm trying to install SD for the first time and I keep getting hit with this error. Any ideas?

Commit hash: efddd05e11d9cc5339a41192457e6ff8ad06ae00
Fetching updates for Stable Diffusion XL...
Checking out commit for Stable Diffusion XL with hash: 45c443b316737a4ab6e40413d7794a7f5657c19f...
error: The following untracked working tree files would be overwritten by checkout:
        .github/workflows/black.yml
        .github/workflows/test-build.yaml
        .github/workflows/test-inference.yml
        .gitignore
        CODEOWNERS
        LICENSE-CODE
        README.md
        assets/000.jpg
        assets/001_with_eval.png
Please move or remove them before you switch branches.
error: The following untracked working tree files would be removed by checkout:
        assets/sv3d.gif
        assets/sv4d.gif
        assets/sv4d_videos/bunnyman.mp4
        assets/sv4d_videos/dolphin.mp4
        assets/sv4d_videos/green_robot.mp4
        assets/sv4d_videos/guppie_v0.mp4
        assets/sv4d_videos/hiphop_parrot.mp4
        assets/sv4d_videos/human5.mp4
        assets/sv4d_videos/human7.mp4
        assets/sv4d_videos/lucia_v000.mp4
        assets/sv4d_videos/monkey.mp4
        assets/sv4d_videos/pistol_v0.mp4
        assets/sv4d_videos/snowboard_v000.mp4
        assets/sv4d_videos/stroller_v000.mp4
        assets/sv4d_videos/test_video1.mp4
        assets/sv4d_videos/test_video2.mp4
        assets/sv4d_videos/train_v0.mp4
        assets/sv4d_videos/wave_hello.mp4
        assets/test_image.png
        assets/tile.gif
        assets/turbo_tile.png
Please move or remove them before you switch branches.
Aborting```

[ornate elk]

Hey, have you checked the install guides from the pinned messages?

[mystic basin]

like this?

to add in the webui-user.bat

i also tried disabling my igpu so i could use automatic1111 but it didn't work so i enabled my igpu back

[worldly trail]

I did - am I missing something obvious?

[ornate elk]

You have weird stable video stuff assets in your webui

[ornate elk]

Oh sry the gfx1031 is your 6850mxt
So your just missing the correct rocm files

You need these files:
https://github.com/likelovewant/ROCmLibs-for-gfx1103-AMD780M-APU/releases/download/v0.6.1.2/rocm.gfx1031.for.hip.sdk.6.1.2.7z

[mystic basin]

So I download these add them to my rocm folder?

[ornate elk]

You have to do this step:
Go into C:\Program Files\AMD\ROCm\6.1\bin\rocblas folder.
There copy and rename the library folder to old_library
Open the .zip file and drag and drop all files of the library folder into the library folder. Not into the old_library folder.
The rocblas.dll goes into the rocblas folder.
Important Restart your PC after this Step before proceeding.

Then relaunch the webui-user.bat

[wraith valley]

does it want bytorch ?

im installing comfiUI

https://youtu.be/T_zUPUU2DdY?si=i_kkQkgGhFMjyl4i&t=224

added same minute in the link

[solemn moon]

is there a step needed for Face Fusion? The interface works and shows updates when changing the frame, just doesnt export or save when i press start..

raceback (most recent call last):███████████████████████████████████████████████████▌| 814/819 [00:07<00:00, 63.15it/s]
File "C:\Users\o0\stable-diffusion-webui\venv\lib\site-packages\gradio\routes.py", line 488, in run_predict
output = await app.get_blocks().process_api(
File "C:\Users\o0\stable-diffusion-webui\venv\lib\site-packages\gradio\blocks.py", line 1431, in process_api
result = await self.call_function(
File "C:\Users\o0\stable-diffusion-webui\venv\lib\site-packages\gradio\blocks.py", line 1103, in call_function
prediction = await anyio.to_thread.run_sync(
File "C:\Users\o0\stable-diffusion-webui\venv\lib\site-packages\anyio\to_thread.py", line 33, in run_sync
return await get_asynclib().run_sync_in_worker_thread(
File "C:\Users\o0\stable-diffusion-webui\venv\lib\site-packages\anyio_backends_asyncio.py", line 877, in run_sync_in_worker_thread
return await future
File "C:\Users\o0\stable-diffusion-webui\venv\lib\site-packages\anyio_backends_asyncio.py", line 807, in run
result = context.run(func, *args)
File "C:\Users\o0\stable-diffusion-webui\venv\lib\site-packages\gradio\utils.py", line 707, in wrapper
response = f(*args, **kwargs)
File "C:\Users\o0\stable-diffusion-webui\extensions\sd-webui-facefusion\facefusion\uis\components\output.py", line 51, in start
conditional_process()
File "C:\Users\o0\stable-diffusion-webui\extensions\sd-webui-facefusion\facefusion\core.py", line 195, in conditional_process
process_video()
File "C:\Users\o0\stable-diffusion-webui\extensions\sd-webui-facefusion\facefusion\core.py", line 219, in process_video
if predict_video(facefusion.globals.target_path):
File "C:\Users\o0\stable-diffusion-webui\extensions\sd-webui-facefusion\facefusion\predictor.py", line 57, in predict_video
_, probabilities = opennsfw2.predict_video_frames(video_path = video_path, frame_interval = FRAME_INTERVAL)
File "C:\Users\o0\stable-diffusion-webui\venv\lib\site-packages\opennsfw2_inference.py", line 236, in predict_video_frames
cv2.destroyAllWindows() # pylint: disable=no-member
cv2.error: OpenCV(4.10.0) D:\a\opencv-python\opencv-python\opencv\modules\highgui\src\window.cpp:1295: error: (-2:Unspecified error) The function is not implemented. Rebuild the library with Windows, GTK+ 2.x or Cocoa support. If you are on Ubuntu or Debian, install libgtk2.0-dev and pkg-config, then re-run cmake or configure script in function 'cvDestroyAllWindows'

[pale magnet]

hi again guys! is there anyone that would be able to help me with this? it doesn't have to be with my own code; i wanted a general script for a text-to-image pipeline that is finetuned to a image folder, where you can set the path. I'm trying myself with a LoRA, but i wasnt able to go far

it's for my college paper so it would be very much appreciated. thanks in advance!

[restive temple]

Anyone have trouble clonig SD3-medium to colab? I keep running into this 'entry not found' error ... looks like its there to me lol.. also the previous versions clone just fine..

[ornate elk]

You need to have ffmpeg and added to the path

[haughty tundra]

hey y'all, rejoining this server momentarily to ask about an issue I've been having

I have a gigabyte rtx 3080 OC gaming rev 2.0, and it seems that whenever I use A1111 to generate images, something happens with the card that causes it to send one of the fans into a horrible 3k rpm state while the other fans go completely dead

I've already sent the card back twice to gigabyte, first they replaced the fan, and it kinda worked, but when I tried doing some generation again, it sent the NEW fan up to that 3k plus rpm range and it would NOT be able to be controlled, no matter the software

is there some sort of module in automatic1111 that overides the fan controller modules on the card? That's the only thing I can really think of

[ornate elk]

Hey, nope never heard of such an issue in the last 2 years here

Have you tried playing a high end Game or making a Benchmark to check if it happens there too?

[haughty tundra]

Well, that's the REALLY weird part

When I play a demanding game

Or put the card under stress

it IMMEDIATELY starts working fine again

zero issues

no crazy fans

just

100% normal behavior

and then sometimes, when I STOP the game, it goes BACK to the crazy fan

[ornate elk]

Hmmm, does the zero fan mode works?

[haughty tundra]

I don’t know what that is

[ornate elk]

If you have nothing running the fans should turn off automatically

If the GPU temp is low enough

[haughty tundra]

Yeah

I have that running with Msi afterburner

[ornate elk]

Have you tried to reset all settings to default in afterburner?
Because afterburner is a software that causes more issues than helping

[haughty tundra]

If anything, afterburner actually helping with the issue sometimes

It was REALLY weird

[ornate elk]

Have your tried a full driver reinstall ?

Did you had issues before using afterburner too?

[indigo dirge]

I downloaded (pytorch) manually, how can I install it?

[ornate elk]

By going in a venv with cmd, then running the activate.bat and then using pip install name of torch.whl

[indigo dirge]

Please write me the words separately.. What do I write in cmd?

[ornate elk]

But if you want to upgrade torch in the stable-diffusion-webui then you can just run git pull and then delete the venv folder

[indigo dirge]

I don't know if I have it or not. It was requested by stable-diffusion. I downloaded it on another computer because my internet is bad.

[ornate elk]

Did you installed it with my guide from the pinned messages?

[ornate elk]

In the scripts folder

Is the activate.bat

[indigo dirge]

I downloaded the file via Download Manager on another computer because I can't download it via cmd because my internet is bad

[ornate elk]

Then you should have downloaded the standalone version

The SD.webui.zip
https://github.com/AUTOMATIC1111/stable-diffusion-webui/releases/tag/v1.0.0-pre
And then run the update.bat
And then the run.bat

[ornate elk]

You need to place the torch whl file into the webui folder.
Then you need to run:
pip install "name if the file.whl"

Without the ""

[indigo dirge]

Thanks for the help.. but he also downloaded something else

[cedar condor]

Hello guys

[ornate elk]

Yea it will install packages that are needed

[timid plaza]

Looking to do an AI generated headshot. I would rather not pay the price for a live professional headshot and there are a number of AI companies out there... But then thinking I might be able to do just as well myself. Anyone working with a workflow which does well in taking an image and enhancing it with minimal changes?

[ornate elk]

It seems you didn't have run the command inside the venv.
You have to first run the other command with the venv activate like in your screenshot before this one

[indigo dirge]

When I do the command inside ven this is what happens

[indigo dirge]

But when I do the command inside webui, it starts downloading a file, but in the end, an error appears.

[ornate elk]

Thats a connection error

You should go somewhere with a better connection for the initial installation of the webui

[mystic basin]

sorry for the late answer, i tried but i doesn't seem to work.

i downloaded the file, i copied my folder library and named the copy old_library then i copied all the file from the 7zip file i downloaded then i pasted them in the library folder gfx1035 isn't in the file you told me to download. so either i download an older version of rocblas since even though my graphic card went out after the rx6800, it's less powerful. or i need to do the same thing you did, tell me to download other file

(sorry for my english)

wait nvm i think i found how to fix it

without the set HIP_VISIBLE_DEVICES=1

it try using my igpu

i forgot to add it back after my first test

i'm so dumb
it seem to work for now, it's compiling and it hasn't shown any error for now

i keep this like that or change it to 1024 and keep 128 , i was told to put 1024 and 128 with directml a few month ago

i don't know if it's because of my pc isn't strong enough or if flux isn't supported by automatic1111 yet but it didn't load the model when i tried loading it

but anyways all the rest seem to work

thanks everyone who helped me

[mystic basin]

nvm it doesn't work

i ain't even using cuda, i have an amg gpu, cpu and igpu

do i need to modify thisi?

i deleted a multi prompting management extension and haven't received the error again so far so i guess i'll wait to see if i'll get it or not

sorry if i'm flooding this chat

[mystic basin]

it worked, everyone who helped me and everyone else have a nice day

[mystic basin]

i've had a few problem with my vram, does putting --medvram in webui-user.bat a good way to avoid the problem the vram problem? i've tried to max out tiled vae and other setting but i doesn't work, 12 gb isn't enough i think

still get error

[mystic basin]

might need to switch to --lowvram

[solemn moon]

installed ffmpeg and added to the path
still get errors though

[mystic basin]

i've tried with multiple thing but never work, it work at the start but exactly at the end it give me a vram error

[solemn moon]

have you checked your drivers are up to date

usually also happens with that

[mystic basin]

no didn't but i'll check

thanks for the info

[mystic basin]

my drivers seem up to date
(i had to do something before checking)

nvm forgot to update my amd adrenalin software thing, i was missing one update

i hope it'll fix the vram problem

[haughty tundra]

yes I did have the issues before afterburner

[ornate elk]

Flux isnt supported in Auto1111 currently

[mystic basin]

oh that's why it didn't work

[ornate elk]

Yea please try with a 1.5 or sdxl model

[mystic basin]

is there any problem with vram when using zluda? i didn't manage to generate any image since i always had vram issue at the end when the picture was almost ready. if it's isn't zluda related then i'll need to retest later when my driver are gonna be updated

[mystic basin]

Automatic1111 support pony right?

[ornate elk]

If you get the vram error again show me the full cmd output

And your txt2img settings.

First try with a small model like Dreamshaper v8 and default settings to get it working

[mystic basin]

alright let me test to see if i get the error again

[ornate elk]

Out of vram can be caused by a to high resolution or by using hires fix with the wrong settings

Also remove the flux model from the models folder if you still have it

[mystic basin]

do i keep the sd3 model i have?

or it isn't supported yet

[ornate elk]

SD3 can work, but it depends. How big is the file size of it?

[mystic basin]

15.3gb

[ornate elk]

Okay thats unlikely to work with your GPU or only very slow

[mystic basin]

so i should remove it?

[ornate elk]

As SD3 isnt that good supported I would remove it too

Your fine with 1.5, SDXL and Pony

[mystic basin]

alright i'll remove it after my image is generated if i don't have the vram issue

50% of the image done

wait nvmm it got stuck at 48% and ran out of vram

[ornate elk]

Was that with SD3 ?

[mystic basin]

no

[ornate elk]

Can you show your txt2img settings?

Like resolution etc

Or was highres fix enabled?

[mystic basin]

highres fix was enabled

[ornate elk]

Which model did you used?

1.5 or sdxl/pony

I'm AFK, deving to work. Brb

[mystic basin]

pony i think

[mystic basin]

alright

[ornate elk]

Try add --medvram-sdxl to the webui-user.bat and try again

[mystic basin]

I’ve tried —medvram earlier but didn’t seem to fix it, I don’t know the difference between —medvram and —medvram-sdxl but I’ll try

if it make --medvram only work with sdxl model then i will need to try with an sdxl model since i got vram error again

wait not enough space for all the text

File "C:\SD-Zluda\stable-diffusion-webui-amdgpu\modules\lowvram.py", line 70, in first_stage_model_encode_wrap
return first_stage_model_encode(x)
File "C:\SD-Zluda\stable-diffusion-webui-amdgpu\repositories\generative-models\sgm\models\autoencoder.py", line 321, in encode
return super().encode(x).sample()
File "C:\SD-Zluda\stable-diffusion-webui-amdgpu\repositories\generative-models\sgm\models\autoencoder.py", line 308, in encode
h = self.encoder(x)
File "C:\SD-Zluda\stable-diffusion-webui-amdgpu\venv\lib\site-packages\torch\nn\modules\module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "C:\SD-Zluda\stable-diffusion-webui-amdgpu\venv\lib\site-packages\torch\nn\modules\module.py", line 1541, in _call_impl
return forward_call(*args, **kwargs)
File "C:\SD-Zluda\stable-diffusion-webui-amdgpu\extensions\multidiffusion-upscaler-for-automatic1111\scripts\tilevae.py", line 384, in call
return self.net.original_forward(x)
File "C:\SD-Zluda\stable-diffusion-webui-amdgpu\repositories\generative-models\sgm\modules\diffusionmodules\model.py", line 589, in forward
h = self.mid.attn_1(h)
File "C:\SD-Zluda\stable-diffusion-webui-amdgpu\venv\lib\site-packages\torch\nn\modules\module.py", line 1532, in _wrapped_call_impl
return self._call_impl(*args, **kwargs)
File "C:\SD-Zluda\stable-diffusion-webui-amdgpu\venv\lib\site-packages\torch\nn\modules\module.py", line 1541, in _call_impl
return forward_call(*args, kwargs)
File "C:\SD-Zluda\stable-diffusion-webui-amdgpu\modules\sd_hijack_optimizations.py", line 601, in cross_attention_attnblock_forward
w2 = w1 * (int(c)(-0.5))
torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 3.91 GiB. GPU

i've tried with an sdxl model and vram error

[ornate elk]

Make sure so set the tiled VAE to default settings when trying to upscale (relaunch the webui and dont change the tiled VAE settings. Only enable it)

If not try pony with hires fix disabled on 1024x768

[mystic basin]

i didn't change the tiled vae setting but if i change the hires fix, put it to 1.5 and making the image bigger will this use less vram? i would say thaat it use more but i'm note sure

[mystic basin]

then go in extra to upscale it?

[ornate elk]

You can upscale in img2img with the SD upscale script

[ornate elk]

Only if you use the same resolution as before

Also make sure you only have --medvram-sdxl and not --medvram in the webui-user.bat

[mystic basin]

alright, i'll test without the hires fix to see if it work

i tried pony hires fix and it worked

[ornate elk]

With which settings?

[mystic basin]

1024x768 pony

[ornate elk]

Ah okay

[mystic basin]

thanks for the help

[ornate elk]

No problem.
Later I can show you how to upscale in img2img.

If you use 1.5 models you can use hires fix upscale by 2.
For pony you can try set upscale by 1.5 and use a smaller resolution

[mystic basin]

i'll go sleep for now, it's 2 am for me, have a good night / day

[ornate elk]

Alright, Good night

[indigo dirge]

I use these things but when I make an image the computer becomes very slow and everything stops

is there something wrong؟

is there something wrong 3060 12g
ram 16g

[steep ivy]

Hey, so I got a recommendation the other day about that pony diffusion model, I got the realistic one, but all of the outputs are quite uncanny and weird. I've played around with all the metrics (changed SFG scale, sampler, steps and even different negative embeddings) and even copied 1:1 other people's setttings and prompts and I keep getting these results. Any suggestions? :)

[ornate elk]

you have to increase the windows pagefile if oyu only have 16gb ram

[indigo dirge]

How do I do this?

[ornate elk]

you have to use a larger resolution.
Pony is based on SDXL. Bot hare trained on a native resolution of 1024x1024

so try 1240x1024 for example

[indigo dirge]

Can I buy a bigger RAM? What do you recommend? Will it make a big difference?

[ornate elk]

it will help for using flux and sdxl and also games

32gb is mostly enough

[ornate elk]

here is a quick guide:
https://www.tomshardware.com/news/how-to-manage-virtual-memory-pagefile-windows-10,36929.html

Make sure its only enabled for C drive.
Then set it to 16000Min and 24000Max
Then restart the PC

[indigo dirge]

ok thanks

[ornate elk]

i see you already did that 4 days ago

[ornate elk]

for flux try removing the t5 and clip lencoder and generate again

[indigo dirge]

Yes I did that...so I will buy a bigger and better ram

[steep ivy]

Is 16GB too little for good stable diffusion performance? (I5-8400 & rtx 2080 super)

[ornate elk]

Its not that good for flux but works for sdxl.
If you want a better performance you can increase the Windows Pagefile.

[visual bane]

I am working on generating interior designs from existing room images, but I'm facing an issue where the generated furniture appears too small compared to the room size. I've been experimenting with different control weights for depth and MLDS, but when I set the weights above 0.4, the input image turns out empty. Could you suggest other control types that might work?

Prompt and control-net Setting:

Transform an empty bedroom into a photorealistic, fully furnished interior with high-quality furniture and decor. Ensure the design maintains the room's exact dimensions and architectural features, with a color palette and realistic textures. The composition should be balanced, showcasing professional lighting without structural changes.
Negative prompt: Messed-up room, distorted walls, altered ceiling, changed windows, too many doors, unrealistic sizes, blurry details, low resolution, poor lighting, unfinished appearance, cartoonish or anime style, structural changes, missing features, new wall colours, bad perspective, distorted furniture, oversized decorations, cluttered space, mismatched styles, fake shadows, low quality.
Steps: 20, Sampler: Euler a, Schedule type: Automatic, CFG scale: 13, Seed: 2637739332, Size: 720x360, Model hash: aae64db561, Model: model, Denoising strength: 0.8, ControlNet 0: "Module: mlsd, Model: mls [a78de37a], Weight: 0.3, Resize Mode: Crop and Resize, Processor Res: 512, Threshold A: 0.1, Threshold B: 0.1, Guidance Start: 0.2, Guidance End: 1.0, Pixel Perfect: True, Control Mode: Balanced", ControlNet 1: "Module: depth_anything_v2, Model: control_sd15_depth [fef5e48e], Weight: 0.3, Resize Mode: Crop and Resize, Processor Res: 512, Threshold A: 0.5, Threshold B: 0.5, Guidance Start: 0.0, Guidance End: 1.0, Pixel Perfect: True, Control Mode: Balanced", Version: v1.10.1

[ornate elk]

Try not not using depth but normal map with mlsd

[loud pond]

@ornate elk

all my images are coming out with faces looking incredibly horrible

how do i fix this?

[loud pond]

this is the promt used

[kindred pivot]

Hey there, anyone super good at comfyui ?

[summer osprey]

Brand new to Stable Diffusion and AI/ML in general. Getting comfortable with Python and understanding the AI Ecosystem.

I am trying to run my first model and it seems to be taking extraordinarily long. I have an Intel i9, 4080 GPU and 96gb RAM.

is there a reason my 4080 is only 35C and running at 60w/360w when the model is active and running in Python?

[ornate elk]

Wrong resolution.
Pony is sdxl bases on trained on 1024x1024 resolution

[loud pond]

oh ok

[ornate elk]

Hey, which webui do you use?

[summer osprey]

I am very new to this and haven't worked with any GUI's yet. I used the 'diffusers' library for the stable diffusion img 2 vid. Then downloaded the reference py files (image_processor" and "pipeline_stable_video_diffusion".

[summer osprey]

reply #2. I don't see an obvious channel, but is there a good stable diffusion text to image/video tutorial that works from beginning to end?

[ornate elk]

In the pinned messages of this channel you'll find my step by step setup Guides for different webuis

Its the first link

Its text based but easy to follow.
Start with the Automatic1111 setup.
Thats the common stable diffusion webui

These webuis are all optimised for the best performance and vram/ram usage.
So no need to use the diffusers library

[summer osprey]

So before any worry of why my GPU isn't being utilize, try the webUI first?

[ornate elk]

Yes. Also the GPU utilisation will always be low because the stable diffusion process uses the vram and cuda cores not the 3D acceleration

[summer osprey]

Thanks, and for that kind of knowledge, where do I go to review? I look forward to following your advise and learning.

[main rune]

@ornate elk
got this
Failed to load ZLUDA: Could not find module 'C:\SD-Zluda\stable-diffusion-webui-amdgpu-forge\.zluda\nvrtc64_112_0.dll' (or one of its dependencies). Try using the full path with constructor syntax.
any ideas what I did wrong?

[ornate elk]

There are some sites to learn about the webui usage but I dont have one to recommend rn. You can also ask here or in #📝｜prompting-help

[ornate elk]

What's your GPU?

And do you have the full cmd log?

[loud pond]

if my machine is not powerful enough should i just upscale it?

[main rune]

rx 6600xt

but the odd thing is I can see the file is there

[ornate elk]

You can try generate at 768x768 or near that. That can work for pony too

[loud pond]

oh alright ill try

[ornate elk]

Try delete the .zluda folder from the stable-diffusion-webui folder and relaunch

[main rune]

it works, but is
CUDA Using Stream: False
ok?
Regardless I'm gonna try to generate one image

[signal sigil]

@ornate elk Hey m8. This will probably be a obvious question, so i apologize.
If im training a LoRa on a subject with multiple angles. Do i tag what side the angles are in the captions?
Like side view/front view, etc?

[main rune]

It's...generating something but it's clearly not fully following the prompt and the end result is a grey box

gonna try restarting the webui

[ornate elk]

I think you dont need to specify it as the base model you train on should understand the different angles

[ornate elk]

Does it work now?

[ornate elk]

you can add --cuda-stream to the webui-user.bat to enable it

[hardy canopy]

Any idea why I'm getting grids of the thing I'm generating?

a silly womans faces as a cartoon emoji
Steps: 30, Sampler: DPM++ 2M, Schedule type: Karras, CFG scale: 7, Seed: 4066254049, Size: 1024x1024, Model hash: c9e3e68f89, Model: juggernautXL_v9Rundiffusionphoto2, Version: f2.0.1v1.10.1-previous-527-g720b80da

[main nest]

Hello guys, when I run the bat file in the installation of the stable diffusion im getting "Python not found" but I installed it from her: https://www.python.org/downloads/release/python-31011/

any idea

[ornate elk]

Run the installer again and click, modify, next, and check "add python to environment variable's"

Then relaunch the webui-user.bat

[main nest]

ohhh my friend

I'm DJ_BAN if you remember me

😄

new acc

still the same

[ornate elk]

Uninstall and reinstall python and restart the PC

[main nest]

will do

[ornate elk]

Idk sry xD

[main nest]

done

still the same issue

[ornate elk]

What 😮

[ornate elk]

Can you show me what it looks like when you now run the installer again, click modify, next,

[ornate elk]

Click next

And then screenshot

[ornate elk]

Hmm looks okay

Do you have an antivirus Programm?

[main nest]

Brand new PC

from today, nothing exept drivers

[ornate elk]

Make sure McAfee is not installed

Have you run the python install as admin?

[main nest]

still the same

wanna hop in a call without talking just to ss

[ornate elk]

I'm not at the PC rn

[main nest]

any other suggestions what it could be

ohhh i think I might have an idea

[ornate elk]

You can add the python path to the python.exe Directly into the webui-user.bat

At the PYTHON=

[main nest]

This might be the problem huh

[ornate elk]

Hmm idk if that can cause such issue

[main nest]

how do i do that

[ornate elk]

Right click the webui-user.bat and click on more options, edit

Then you have to go into the python folder

Its in yourUserName/ appdata/local/programs/python

[main nest]

I don't have it there

[ornate elk]

but here it shows:

[main nest]

found it

C:\Users\GRIGS\AppData\Local\Programs\Python\Python31

it was here

and now what do I do 😄

[ornate elk]

it should then look like this:

in the webui-user.bat

aldo whats your gpu?

so you can add the needed performance args ther too

[main nest]

4070 super 12gb

[ornate elk]

then at the line set Commandline_Args= you have to add --xformers --no-half-vae

[main nest]

worked

Perfectly

Thank you, good night and all the best

[ornate elk]

np, gn!

[plain pond]

sorry for bothering. Can u guys help me please, i do all fixes in github, chat gpt but still not working, i try downgrade and reinstalling py,pydantic,albumentation,gradio etc.. to competible version but still this error.., ido --skip instalation too.. and then get error again. before i reset factory windows, i never get problem. the install.. and using au1111.. is never have a problem.
my spec : intel core i5 10300h, rtx 2060 max q 6gb vram, ram 16gb.. thanks.. now i try reinstaling python,git,and automatic111.. thanks.

[astral jasper]

Hey guys, which version of Pytorch should I use to run fooocus with these days?

[main rune]

@ornate elk
I'm only getting grey boxes, I thought it might be loras so I deleted those just to be sure, then I even deleted the VAE when it still does it, erased and redownloaded multiple ckpt and it's still grey boxes

I think I made a mistake installing it so I'm uninstalling/deleting everything again and hope for the best

[ornate elk]

Gray images can be caused by trying to use flux in auto1111 or a lora as model

[ornate elk]

2.3.1 cu12.1

[main rune]

pony, so not that issue

I let the civitui helper extension sort the models for me for the most part

I must've screwed up royally somewhere, I've been trying to just have it try to generate princess peach as a test and it just throws smudges and grey boxes my way

[ornate elk]

What's in your webui-user.bat?

[main rune]

default with --zluda and --cuda-stream in the args

[ornate elk]

Remove --disable-nan-check from the webui-user.bat
Then delete the venv folder and relaunch

[ornate elk]

Can you screenshot your txt2img settings?

[ornate elk]

Is it a model or a lora? Check the file size of it

[main rune]

really no deviations, the most was me trying if 1024x1024 would somehow work

[main rune]

it's definitely a model, I used it before

[ornate elk]

Can you show me the top bar of the Webui?

[ornate elk]

Okay looks normal

What does the cmd shows and is the image pure Gray?

[main rune]

pure gray

and no errors

wait

at least not ones visible to me

[ornate elk]

Strange

[main rune]

I think I might've messed up the zluda files or something

[ornate elk]

You could try install Automatic1111 and see if that works

As you already have zluda running, its quick to test

[main rune]

I'm using forge because automatic1111 has never worked properly for me
I had forge running before but it suddenly borked on me so I reinstalled everything

[ornate elk]

Ah okay

[main rune]

I'm beginning to think that maybe my hardware is dying, but it is trying to output something so I'll test more

[ornate elk]

I dont think so

Also did you installed hip SDK 5 or 6 ?

I will update my guide for the lower cards for hip SDK 6 tomorrow maybe

[main rune]

huh, I forgot, I think I installed 5 but it's possible I might've installed 6 as well while following other guides
don't remember if I uninstalled those

[ornate elk]

Okay, maybe recheck that.

I need to go off now. Can help more tomorrow

[main rune]

thanks

Gn

[ornate elk]

Gn

[knotty turret]

@ornate elk Need help cant find this custom node

now say its id 170 but i already have it

[plain pond]

Thanks man that work but i got new problem.. actualy this is one of the first problem ive head after reset windows..

[astral jasper]

So I have simple question, what is the best option to run fooocus when I can't use pytorch with jupyter? Thinking about 7777 port but not have clue how to use it 😅

[main rune]

@ornate elk if you're awake, or if anyone else knows, is the zluda guide for forge specifically for the 5.7 HIP SDK?

[ornate elk]

I made if for both. Newer cards on 6 and older ones with 5.
But yours should work with 6 too

[ornate elk]

How much Ram do you have?

[uncut heron]

where can i find - ImageResize+_legacy-0283ce4 - for ConfiUI?

[solemn moon]

GPT says

https://github.com/cubiq/ComfyUI_essentials

[plain pond]

16 sir.. this is my spec : intel core i5 10300h, rtx 2060 max q 6gb vram, ram 16gb

[charred rover]

xformers 0.0.20 requires torch==2.0.1, but you have torch 2.4.1
and if I install 2.4.1 I cant install xformers, incompatible

(CUDA 12.5)

actually CUDA 12.4 is installed too, should I delete it?

[ornate elk]

Then you need to increase the Windows Pagefile

[charred rover]

when trying to install xformers:

Traceback (most recent call last):
File "<string>", line 2, in <module>
File "<pip-setuptools-caller>", line 34, in <module>
File "C:\Users\HYPERPC\AppData\Local\Temp\pip-install-a38qitp\xformers_d438b9c1cf214cc3bf1021dfcb8611b8\setup.py", line 24, in <module>
import torch
File "C:\Users\HYPERPC\AppData\Local\Programs\Python\Python310\lib\site-packages\torch_init.py", line 137, in <module>
raise err
OSError: [WinError 127] The specified procedure could not be found. Error loading "C:\Users\HYPERPC\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\lib\c10_cuda.dll" or one of its dependencies.

I do have this file c10_cuda.dll

'Usually happens when PyTorch and CUDA versions are incompatible, or when there are mismatched libraries (like torch and xformers) on your system.'

[plain pond]

thank you very much.. thats work.. but i think the problem won't stop here haha.. idk why i just install control net.. but this is happend

[azure magnet]

But if i have 2 gpu with 12gb, will they work as one with 24?

[vocal burrow]

nope can't happen, best you can do is run two instances, each using 12gb

[vocal burrow]

full log ?

[median geode]

hello, Can u guys help me please, what's wrong with my wd14 tagger

and also why is there AMD in the second line? mine is nvidia

[vocal burrow]

Network connection error, if you live in """certain countries""" you might have to use a vpn. If not it could be very bad wifi.

[vocal burrow]

AMD64 refers to your cpu architecture being 64bits, nothing to do with your cpu or gpu manafacturer.

[median geode]

hahaha """certain countries""" .ok fine thanksXD

[uncut heron]

Yeap i used ChatGPT
Still wasnt able to understand

[ornate elk]

try delete the venv folder and relaunch

[loud pond]

@ornate elk any way to make a 1024x1024 image with low vram? Do I need any plugins or something?

[rustic wasp]

my lora folder on the gui has been showing 2 tabs and showing the same lora twice on only the first page, i have no idea why is this happening and it keeps showing up even on fresh installs of lora's, this is on webui forge, anyone know how to fix this? it hasn't given me any errors it's just annoying

[rustic wasp]

not the one you asked for but you could try command arguments, my pc also has low vram

these are mine "--lowvram --no-half --opt-split-attention"

but you could look here for more info on they do

https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Command-Line-Arguments-and-Settings

[loud pond]

thanks! ill give it a try

[ornate elk]

use the tiled vae extension and --lowvram

also try add --opt-sub-quad-attention

[loud pond]

for the tiled vae extension what should i use the settings as?

[ornate elk]

thats good

[plain pond]

Thats work, Thank you very much for helping me out.. Gn sir

[ornate elk]

no problem gn 🙂

[loud pond]

It's still giving me the not enough bytes error, do I have to lower the settings?

[ornate elk]

which model do you try?

try if 768x768 works