🤝｜tech-support

[ornate elk]

also whats in your webui-user.bat?

[dim fractal]

doing that now

[dim fractal]

@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

normal sD forge ui

[ornate elk]

oh your usinf forge. thats outdated and not recommended

switch to Auto1111 and you shouldnt have any issues

[dim fractal]

have to use forge, vram issues

worked fine until 3 days ago when i did my wipe

for 6 months

[ornate elk]

a rtx4060 works good in auto1111 you just need to use the right commandline args

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

[dim fractal]

believe me when i say.. forge is around 45% faster generation, which is very useful for me creating work for commissions, i can make $2400/mo with forge, with 1111 i can make around $1400-1600

something to do with my enviroment paths, cudnn, cuda, onxxx, but idk, looks fine to me, i cant figure it out

[ornate elk]

we had some people here which told me the same but we could make auto1111 work only maybe 5-10 seconds slower.
Here is the Forge Announcement. So thats why its not recommended for normal Users anymore. There is a fork of Forge maybe you should use that instead if you dont want Auto1111.
https://github.com/lllyasviel/stable-diffusion-webui-forge/discussions/801

[dim fractal]

im using the fork, yes

[ornate elk]

ahhh okay

[ornate elk]

in reactor you can set it to cpu mode

isnt much slower if cuda wont work as its only for face detection/Swap

[dim fractal]

you can? oh dang, i didnt know that

uh ill go see to figure out how

[ornate elk]

yea its not in the webui settings but in the reactor tab itself

under reactors settings

[dim fractal]

will try, thanks!

[dim fractal]

works fine 🙂

and yeah its not slow at all, quite fast actually

[ornate elk]

perfect! 🙂

[tame yoke]

can i run txt2vid with nvidia gtx 1650

[ornate elk]

I updated the guide.
Fooocus should work again

[dense tapir]

Where do I change the default prompts, CFG, Seed, Adetailer settings, etc...?

[misty marten]

My inpaint does not fill the selected part, I've tried several models and changed several options, but it still gives the same results. my prompt for this image was wearing medieval armor.

[dense tapir]

Try 1024x1024 and tick Fill

Set denoising to 0.59 and slowly go up if it's still bad

[misty marten]

It didn't work, I still have the same results :/

[dense tapir]

Try Just Resize

[fathom rose]

please help me to set up

[lost chasm]

hey dumb question, 4090 comes with a power connector that feeds in 4 power cables from PSU but this PC i'm trying to ram it in only has 3 to feed into it
will it work with 3 or is this a dumb question, asked by an idiot

[ruby yoke]

i wouldn't do that if i were you

[lost chasm]

i think its a great idea what could go wrong?

[ruby yoke]

i can think of a lot of things that can go wrong. might do a little bit of research on google

[lost chasm]

nah too late, plugged it in, it works, woohoo! party time.

[ruby yoke]

do i smell smoke?

[lost chasm]

nein

this is how tough guys do it breh

no research

[ruby yoke]

knew a guy years ago that thought like that. he had a vga monitor, the card in the computer was cga. he was sure all he had to do was put extra pins in the monitor plug. used some bits of wire to do that. bought new equipment after

[lost chasm]

yeah but i skimmed some reddit forums or whatever

and once i read something that sounded like what i wanted to hear, that was good enough for me

[ruby yoke]

there's a reason there are 4 plugs, you know?

[lost chasm]

yeah for overclocking

[ruby yoke]

do you also cut the ground plug off of extension cords?

[lost chasm]

sure, i do all sorts of cool things

[ruby yoke]

it's your money

[lost chasm]

i emailed Falcon Northwest support but they are too slow, tough guys never wait.

if it makes you feel better, i wont come here and complain to you if i short out my gpu lol

[dense tapir]

Chad

[lost chasm]

i got the vibe it would work fine and i factchecked my vibe with my emotions and my emotions gave me a good feeling about it, technically.

[dense tapir]

Do tough guys care about emotions?

[lost chasm]

even tough guys cry sometimes

but then they wipe their tears away and punch things

[dense tapir]

About tough guy things

[lost chasm]

https://www.youtube.com/watch?v=j7OHG7tHrNM

ok i did some SDXL gens to give it a spin and nothing exploded so far

https://www.tiktok.com/@camer4obscura/video/7098478422967487749

[fathom rose]

plzz help...

[lost chasm]

"3 pcie 8 pin cables from power supply to included 4090 power adapter"

sounds like i'm golden

a special magical golden idiot who stumbled blindly on the correct solution before i properly confirmed it, like heroes do

[violet void]

Thank you for using comcom analytics.
"comcom analytics" supports all community managers (moderators and server owners) by stats, visualization, and analytics.

If you have any questions, feel free to ask us!
Your dashboard
Help
Support server

Other languages
en: help
ja: help Japanese

[worn bane]

What is this model name?

[vernal wolf]

Im not sure but probably "autismmixSDXL_autismmixPony", in case its sdxl

[worn bane]

All the artist's pictures are in the same style, I don't think sdxl

Search artkoikoi in pixiv

[vernal wolf]

🤔

[worn bane]

search and see the all post style

[vernal wolf]

Oof and I dont remember sd 1.5 model´s names, sorry U.U @worn bane

[worn bane]

Its same style

[worn bane]

💔

[vernal wolf]

Don´t worry, I found it!

[worn bane]

Damn

[vernal wolf]

Its Anythingv3 model based on sd 1.5 @worn bane and the user has a civitai page where he/she uploads the lora models used https://civitai.com/user/artkoikoi

[worn bane]

Hmmm can you make a image using his lora?

[vernal wolf]

Yes but idk which one, there are many of them

If you have sd in your PC you can download the anythingv3 model and try any of the loras, I dont have that model downloaded sadly so i cannot test it rn

[worn bane]

I don't have pc

[vernal wolf]

On civitAI you can use some of the loras so you can generate the images there

[umbral ingot]

Ok so I installed SD and everything and I have a Nvidia 2070 yet when I try to generate anything with default settings and size of the image, it slows my PC down so much it might as well be stuck frozen

Even with the only thing being opened is SD it still struggles for a long time somehow.

[ornate elk]

Make sure Wallpaper Engine is closed/disabled (not paused) too, if you have that.
Then you also need to edit the webui-user.bat (right click)
And at the line commandline_args=
You need to add: --xformers --medvram-sdxl --no-half-vae
Then save and relaunch the webui-user.bat

[worn bane]

LoL i found his settings

[glass sable]

Since the update I've discovered the source of the slowness. "No module found xformers" what ver should I have?

And what's my pip command for it?

A11, obviously.

"set COMMANDLINE_ARGS= --no-half --reinstall-torch --xformers"

[fathom rose]

@ornate elk plz help me

[fathom rose]

i think i figured it but i'm trying to set directory for the checkpoints through command line args or internal setting
Not sure how..

i'm using the same webui-user setting on 1.10 RC as the old versions with the same directory commands but it's not picking up

ahh nvm i think i missed something to copy

now got it fixed

is this a ok setting for the 1.10RC? do i need to change anything here?

[ornate elk]

You should add --xformers for better performance

And remove --medvram

[fathom rose]

adding xformers will allow me to get the full potential of SD3 still?

[fathom rose]

the --medvram-sdxl necessary or no?

my gpu is 3070ti

[dense tapir]

I downloaded a wildcard zip, where do I put the content of it?

Figured it out!

Goes into extensions - sd-dynamic-prompts - wildcards

[fathom rose]

will i need to get the text encoders separately or they come packed with the medium checkpoints?

[glass sable]

"xFormers wasn't build with CUDA support" WTFITS?

I've def got the right ver of xformers for 11.8

Windows, A11, Somebody come to my house and put me out of my misery. That's what this chat is for right?

Here's the error I'm getting front-end: NotImplementedError: No operator found for memory_efficient_attention_forward with inputs: query : shape=(16, 5600, 10, 64) (torch.float32) key : shape=(16, 5600, 10, 64) (torch.float32) value : shape=(16, 5600, 10, 64) (torch.float32) attn_bias : <class 'NoneType'> p : 0.0 flshattF is not supported because: xFormers wasn't build with CUDA support dtype=torch.float32 (supported: {torch.float16, torch.bfloat16}) Operator wasn't built - see python -m xformers.info for more info tritonflashattF is not supported because: xFormers wasn't build with CUDA support dtype=torch.float32 (supported: {torch.float16, torch.bfloat16}) Operator wasn't built - see python -m xformers.info for more info triton is not available requires A100 GPU cutlassF is not supported because: xFormers wasn't build with CUDA support Operator wasn't built - see python -m xformers.info for more info smallkF is not supported because: xFormers wasn't build with CUDA support max(query.shape[-1] != value.shape[-1]) > 32 Operator wasn't built - see python -m xformers.info for more info unsupported embed per head: 64

[ornate elk]

Yes Its needed

[ornate elk]

What's your GPU?

[glass sable]

RTX4090

I'm currently reinstalling Torch.

Bout to check CUDA ver. I have no foqqin clue what I've wrought.

[azure bough]

nvcc -V

it should kinda liik like this

[glass sable]

Same error. Just stepped down to the pyver demanded by the startup command line

I've got 2.4.0+cu118 AND Python 3.11.3, and CUDA 11.8

[glass sable]

nvcc fatal : No input files specified; use option --help for more information

[glass sable]

Reinstalled CUDA. Some muthfakkin error. Please advise.

[ornate elk]

You need python 3.10.11 64bit

Uninstall 3.11
Then install 3.10.11 64bit
Then make sure you only have --xformers in the webui-user.bat commandline_args.
Then delete the venv folder and relaunch the webui-user.bat

[glass sable]

Doing so. Why does it say it requires 3.11.3?

[azure bough]

so download the exe

[glass sable]

Way ahead of you, sibling of indeterminate gender.

[azure bough]

is it solved?

[glass sable]

Noop.

[azure bough]

I initially tried installing CUDA through the command line but encountered issues.
I switched to the executable installer and the installation was successful.

like 1h ago

[glass sable]

You're running A11 on 12.5?

[azure bough]

a1111 and comfyui

wait

12.5?

[glass sable]

Waiting.

[azure bough]

whats 12.5

[fierce hollow]

does anything in SD generation affect SSD write/read speed?

[glass sable]

CUDA 12.5. I was under the impression A11 had to be run on 11.8.

[azure bough]

it dosent matter

[glass sable]

Well, then I'm at a complete loss.

[azure bough]

so you tried the exe right?

what are you doing exactly?

installing a1111?

[glass sable]

I had to rebuild and webui.bat wiped torch, but didn't reinstall it. Everything has been jank since then.

[azure bough]

why dont you reinstall

[glass sable]

What torch build are you running?

I don't wanna lose my setup and porting everything over is a pain in my fat ass.

[azure bough]

listen. Do you want to install a1111 or comfyui.

[glass sable]

I'm just trying to get A11 running on Xformers optimizaton again.

[ornate elk]

It doesn't

It should work fine with 3.10.11

[ornate elk]

You don't need to install any cuda stuff manually

[glass sable]

Unfortunately, it is not.

[azure bough]

if you want to install comfyui then just do iex (irm comfy.tc.ht) do this

[ornate elk]

Can you show me the full cmd log?

[glass sable]

Comfy's working just fine, thanks. It's just shite for inpainting.

[azure bough]

well i am new to sd

[glass sable]

Theygo.

[ornate elk]

You didn't followed my advise from above

Its still using python 3.11

Uninstall 3.11
Install 3.10.11
Delete the venv folder
Relaunch the webui-user.bat

[glass sable]

Sorry.

Forgot to reopen my environment.

1 sec while I do so.

I'm in a 3.10.11 venv, but it's still foqqing showing me at 3.11.5

[ornate elk]

Have you deleted the venv folder?

[glass sable]

No?

Should I be wearing a duncecap?

[ornate elk]

I said twice that you need to delete the venv folder and then relaunch the webui-user.bat

If not it will never use the right Python version

[glass sable]

Duncecap time.

Huh. Now it's on 3.9.13. Thank you for your patience. It is 4 am and I know I am as frustrating as I am frustrated.

[ornate elk]

Then you need to uninstall 3.9

Then delete the venv folder again

[glass sable]

K. Did that. Rerun webui-user?

Traceback (most recent call last):
File "C:\stable-diffusion-webui\launch.py", line 48, in <module>
main()
File "C:\stable-diffusion-webui\launch.py", line 39, in main
prepare_environment()
File "C:\stable-diffusion-webui\modules\launch_utils.py", line 380, in prepare_environment
run(f'"{python}" -m {torch_command}', "Installing torch and torchvision", "Couldn't install torch", live=True)
File "C:\stable-diffusion-webui\modules\launch_utils.py", line 115, in run
raise RuntimeError("\n".join(error_bits))
RuntimeError: Couldn't install torch.
Command: "C:\stable-diffusion-webui\venv\Scripts\python.exe" -m pip install torch==2.1.2 torchvision==0.16.2 --extra-index-url https://download.pytorch.org/whl/cu121
Error code: 1

[ornate elk]

The whole cmd log pls

[glass sable]

I closed it. It's running again. 1 sec.

venv "C:\stable-diffusion-webui\venv\Scripts\Python.exe"
Python 3.10.11 (tags/v3.10.11:7d4cc5a, Apr 5 2023, 00:38:17) [MSC v.1929 64 bit (AMD64)]
Version: v1.6.0-1398-gfeee37d7
Commit hash: feee37d75f1b168768014e4634dcb156ee649c05
Installing torch and torchvision
Looking in indexes: https://pypi.org/simple, https://download.pytorch.org/whl/cu121
Collecting torch==2.1.2
Downloading https://download.pytorch.org/whl/cu121/torch-2.1.2%2Bcu121-cp310-cp310-win_amd64.whl (2473.9 MB)
-------------------- ------------------- 1.2/2.5 GB 28.5 MB/s eta 0:00:44ERROR: Could not install packages due to an OSError: [Errno 28] No space left on device

 -------------------- ------------------- 1.2/2.5 GB 28.5 MB/s eta 0:00:44

[notice] A new release of pip is available: 23.0.1 -> 24.1.2
[notice] To update, run: C:\stable-diffusion-webui\venv\Scripts\python.exe -m pip install --upgrade pip
Traceback (most recent call last):
File "C:\stable-diffusion-webui\launch.py", line 48, in <module>
main()
File "C:\stable-diffusion-webui\launch.py", line 39, in main
prepare_environment()
File "C:\stable-diffusion-webui\modules\launch_utils.py", line 380, in prepare_environment
run(f'"{python}" -m {torch_command}', "Installing torch and torchvision", "Couldn't install torch", live=True)
File "C:\stable-diffusion-webui\modules\launch_utils.py", line 115, in run
raise RuntimeError("\n".join(error_bits))
RuntimeError: Couldn't install torch.
Command: "C:\stable-diffusion-webui\venv\Scripts\python.exe" -m pip install torch==2.1.2 torchvision==0.16.2 --extra-index-url https://download.pytorch.org/whl/cu121
Error code: 1

[ornate elk]

Your using an old webui version

[glass sable]

I just upgraded?

That's what kicked off this very long and painful experience.

[ornate elk]

1.6.0 from august last year

Is your version

Were on 1.9.4

[ornate elk]

Its not upgraded. You need to add
Git pull
Into a empty row into the webui-user.bat

Then delete the venv folder again

And relaunch the webui-user.bat

[ornate elk]

Ohhh and this error is caused because your C Drive is full

So make sure you have 10-15 GB free space on C

[glass sable]

You are a god among men, CS1o

[azure bough]

its fixed? @glass sable

[glass sable]

Optimum is not installed! Installing...
Installing optimum
Launching Web UI with arguments: --xformers
Traceback (most recent call last):
File "C:\stable-diffusion-webui\launch.py", line 48, in <module>
main()
File "C:\stable-diffusion-webui\launch.py", line 44, in main
start()
File "C:\stable-diffusion-webui\modules\launch_utils.py", line 465, in start
import webui
File "C:\stable-diffusion-webui\webui.py", line 13, in <module>
initialize.imports()
File "C:\stable-diffusion-webui\modules\initialize.py", line 23, in imports
import gradio # noqa: F401
File "C:\stable-diffusion-webui\venv\lib\site-packages\gradio_init_.py", line 3, in <module>
import gradio.components as components
File "C:\stable-diffusion-webui\venv\lib\site-packages\gradio\components_init_.py", line 1, in <module>
from gradio.components.annotated_image import AnnotatedImage
File "C:\stable-diffusion-webui\venv\lib\site-packages\gradio\components\annotated_image.py", line 13, in <module>
from gradio.components.base import IOComponent, Keywords
File "C:\stable-diffusion-webui\venv\lib\site-packages\gradio\components\base.py", line 20, in <module>
from fastapi import UploadFile
File "C:\stable-diffusion-webui\venv\lib\site-packages\fastapi_init.py", line 7, in <module>
from .applications import FastAPI as FastAPI
File "C:\stable-diffusion-webui\venv\lib\site-packages\fastapi\applications.py", line 15, in <module>
from fastapi import routing
File "C:\stable-diffusion-webui\venv\lib\site-packages\fastapi\routing.py", line 22, in <module>
from fastapi import params
File "C:\stable-diffusion-webui\venv\lib\site-packages\fastapi\params.py", line 4, in <module>
from pydantic.fields import FieldInfo, Undefined
ImportError: cannot import name 'Undefined' from 'pydantic.fields' (C:\stable-diffusion-webui\venv\lib\site-packages\pydantic\fields.py)

F* Me.

git: 'Pull' is not a git command. See 'git --help'.

[azure bough]

I have all my checkpoints in the comfyui folder, but I want to access them from a1111. What do I do?

[ornate elk]

Can you show the full cmd log again?

I guess its still not on the latest version

[ornate elk]

You can edit the webui-user.bat of auto1111
In the commandline_args line you can add --ckpt-dir "D:\PathToModelsFromComfy
And --lora-dire "..."

[azure bough]

with " ?

[glass sable]

It's working. I have wrk to do. Will update later.

[azure bough]

Thanks

[ornate elk]

No problem

[glass sable]

Odd. Roop is now not showing up in my lower tray. Any idea what I can do? It's showing under extensions...

[azure bough]

I am trying to update from comfyui manager. But doesn't work

[ornate elk]

Roop is outdated and you should switch to Reactor (extension) or Roop-unleashed (seperate tool)

[ornate elk]

You need to run the command it tells you there

It seems like you moved comfyui before

[azure bough]

yea

i did put the commend before but nothing happend

[ornate elk]

You need to do it in a cmd thats opened inside the comfyui folder

[azure bough]

did

[ornate elk]

Then you need to do it again

It mostly shows up to 10 different paths

So you have to do it a lot

[azure bough]

i did it again

[ornate elk]

Hmm, than an other solution is a clean reinstall on the drive you want to have it

Also the drive needs to be formated as GPT and not MBR

Because wrong formated drives dont have an ownership management

[azure bough]

how do check

[ornate elk]

In the diskmanager

Partitionmanager

[ornate elk]

Oh wait

In the partition manager Right click the Y drive

Then on Properties

Then click on Volumes

And there you see it at Partition style

[azure bough]

where is volumes

[ornate elk]

Not there

I meant in the screenshot before

In partition manager

[ornate elk]

Hmm okay

Dont have time rn. Sry but you have to google "how to check if a drive is GPT or MBR"

[azure bough]

ok thanks. bye

[umbral ingot]

Huh, that's oddly specific, I am using Wallpaper Engine. Is there a major complication with WE and SD?

[umbral ingot]

I have --xformers
but not the other two things

[ornate elk]

Yes. Running both even WE on pause can cause PC freezes and even PC crashes. At best it only slows down the generation speed.
So make sure WE is completely closed.

[ornate elk]

Add the other two for better sdxl speed

[umbral ingot]

Got it. I'll try doing that when I get off work

[azure bough]

Never happened to me before

[ornate elk]

Happens only when going over the vram limit, for example using large resolution + upscale

Or sdxl with 8gb vram or less
For the best performance keep it disabled

[azure bough]

ok, I will keep that in mind

[glass sable]

Finally got A11 *trying * to git pull for an update.
error: Pulling is not possible because you have unmerged files.

[ornate elk]

What files?

[glass sable]

It didn't say, but I was able to hunt them down on the interrim. Now I just need to figure out how to get the oss2 module.

[karmic crown]

what is your version of a1111?

show me git log

when you have a merge conflict, its fastest and easiest to restore any older commit and then just do update

for example

git reset --hard adadb4e3c7382bf3e4f7519126cd6c70f4f8557b && git pull

but this will work only if your version is at least 1.9.0

that's why I need to see what you have first

show me the result of git log

for older version, i need to find some older commit

you can use this one without checking:

git reset --hard 20ae71faa8ef035c31aa3a410b707d792c8203a3 && git pull

that is 1.3.0, you should have newer version (i hope) 🙂

[knotty merlin]

This has happened every time I've tried to load a new model since I moved A1111 to my thumb drive.

When I try to reload the model it allows me to but all the results are noticeably lower quality

[void burrow]

is this the place i can ask abt forge ui questions?

[ruby yoke]

move it back then

[gray matrix]

Hi guys, I'm trying to build a "Try on clothes" with Stability's AI API. But not success.
Does Anyone know how to work that?

Sending initial image, mask, and product's image (clothes) to change.
How this example: https://huggingface.co/yisol/IDM-VTON

Thank you for any help!!

[solemn saddle]

Hey guys, not sure if i've come to the right spot but i have a bit of an issue, because i have to regularly use a VPN for work (and sometimes i render images in the background) I was wondering if anyone knows how to set Automatic1111 to launch the webui eventhough a VPN is open, this is how far it gets to with my VPN active, and if i launch the webui then start the VPN i just get spammed with "timeout" errors

before anyone asks, this is my personal rig as i work from home

[upper iron]

hello ive been trying to get the kohya_ss lora to work for training but I always get this error and im not sure what to do, some help would be awesome!

[delicate citrus]

exist diference between install sd on nvdia and amd, or the process is the same?

[upper iron]

Idk im new to all this

[misty marten]

nothing changed, just the colors became saturated :/

[ornate elk]

Its different. For AMD install guides checkout the link from the pinned messages of this channel

[ornate elk]

Are you on AMD ?

[misty marten]

Yes

[ornate elk]

What's your GPU?

It seems your using directml.
To make Inpainting work with it you need to add --no-half to the webui-user.bat.

[misty marten]

RX 6600

[misty marten]

i will try it

[ornate elk]

That card is supported by zluda a faster webui Backend with better vram management than directml.
You find an install guide for it in the link in the pinned messages

[misty marten]

now it worked, but the generated image does not fit the original

[misty marten]

I'm installing it right now

[upper iron]

Guys i need help pleasseee

[misty marten]

zluda is giving this error

[umbral ingot]

Ok yea, that was definitely the reason why it was so slow. It's significantly faster now. Thanks so much

[umbral ingot]

why is it that after a few generations it just gets stuck at 98% when in the cmd window it says its 100%

[ruby yoke]

look in your output folder and see if the image is in it

[umbral ingot]

yea its in there but it still gets stuck at 98% and i cant generate anything until i restart it

[ruby yoke]

if the image is there, then i'm going to suspect something on your system. how much disk space do you have, to start with. and how much memory?

[umbral ingot]

disk space on what my hard drive my directory is in or my c drive? I also have 64GB memory

[fierce hollow]

I reinstalled windows, but saved the folder with SD, what else do I need to install to use SD?

[ruby yoke]

both, as comfy (by default) uses space on both

[ruby yoke]

cuda? python?

[umbral ingot]

well with my c drive im always fighting for more space for some reason. my directory is in my hard drive and it has 8TB of space in total

[ruby yoke]

start clearing stuff on C: out of temp files, and out of the /usr/directories. when you uninstall stuff, thre's a lot of stuff left behind. i'll usually get a copy of Ccleaner and have it look for stuff that's been orphaned in the registry, too

[umbral ingot]

isnt ccleaner spyware?

[ruby yoke]

good lord no

[umbral ingot]

well ive already been deleting my temp folders because of this issue

[ruby yoke]

go into C:\Users and then into the folder with your account name on it

and then go through all the folders under that - you'll find a lot of stuff, with a lot of stuff in them, for stuff you uninstalled.

[umbral ingot]

oh im sure theres a ton of stuff there i can delete but im worried about deleting something that shouldnt be deleted

[ruby yoke]

you didn't read what I said. in the past, as you have uninstalled stuff, folders with files have been left in C:\Users<your account> - go through the folders in there

then go look at your actual drive. you'll find folders, with content, there as well - for stuff that you uninstalled and that you thought ws gone

and then make sure stuff like images and videos are on some other drive

[umbral ingot]

oh, ive already deleted a bunch of stuff in there

[fierce hollow]

I installed python, git, I'm trying to run SD, that's what it gives me, and there's no further progress. It freezes..

[fierce hollow]

help....

[ornate elk]

Can you show me the whole cmd log?

[ornate elk]

Its literally malware

Dont install ccleaner

[ruby yoke]

sigh. no. it isn't. never has been, never will be

[ornate elk]

In the past it was good but now its not even necessary to use it.
User just break their registry with it.
Totally not worth to ever use that programm.
It won't fix anything

[ruby yoke]

yes it does. i use it all the time. have since the first iteration of it

[ornate elk]

Ccleaner got bought by avast and avast got bought by Norton360.
Norton360 install cryptominers with their AV Software.
I wouldnt install anything from these companys on my system

Closed source and mostly spyware now

[ruby yoke]

i haven't installed norton since the dos days. avast used to be good. Ccleaner doesn't have the issues those have, however

[ornate elk]

You don't know what it does in the background

[ruby yoke]

and it's not a boot sector virus like norton

[ornate elk]

It belongs to the same company's so one update can change it

[ruby yoke]

i think you have a scammer in general

[ornate elk]

The team will check it. Maybe we need to include "shady offerings" in the rules

[ornate elk]

Make sure you whitelist the webui in any browser adblocker.
Sometimes they can cause such issues.

[ruby yoke]

yeah, that would be good

[ornate elk]

In 2018 it got hacked and delivered malware to the users.
Also in 2018 they updated it and since then it spy's on the users like avast does too.
There are articles to read online about both incidents.

So don't tell me ccleaner is a good Programm xD

[ruby yoke]

windows spys harder than any of those. and thsi isn't 2018 ... i've had no issues with it

[ornate elk]

Yea but why recommend spyware to users from shady companys

[ruby yoke]

it's not spyware

and there's nothing you can install that doesn't track you - though comfy might not. yet.

[ornate elk]

It does collect user data

Yea comfyui custom nodes is a problem

[ruby yoke]

so does every other program you put on your phone or computer

including your os

the idea of privacy is a myth

[ornate elk]

I'm using GrapeheneOS on Android without Google serviced. Mostly using open source apps behind a firewall.

So my phone spy's less than ccleaner

[ruby yoke]

you're still being tracked, unless you never use the gps functions

and never make a phone call

do either of those, you're tracked

[ornate elk]

GPS and phonecalls don't scan my device for meta data or try to get my browser history or cookies...

[ruby yoke]

that you know of and does that matter? who cares about that, anyway, other than marketing people?

[ornate elk]

Governments like data too

Take a quick look to China, Iran, Russia or myanmar

[ruby yoke]

you use the internet - everything you post goes through military gateways.

and they don't care about your browser history or cookies

[ornate elk]

True but me and other users just don't want more Programms to spy on us.

[ruby yoke]

they already have all the data on you that they need - and they know where you are at all times becaue your phone talks to the cell network

the people you should be afraid of knowing what you're doing and where you are, already do

and the others don't matter

[ornate elk]

If I can minimize data collection I'm fine.
You can't be 100% anonymus. But thats not the goal

[ruby yoke]

don't use google rewards then ...

even though it's realy nice to get the payouts

[ornate elk]

I don't use any google service on my phone

[ruby yoke]

on a totaly different subject, what's happened to lykon?

[fierce hollow]

😭

[ornate elk]

Idk I've not seen him around lately

[ruby yoke]

me either. hope he didn't come down with covid or something

[ruby yoke]

i don't know why you're getting this issue 😦

[fierce hollow]

I hope at least someone will help me...

[ornate elk]

Did you moved your webui ?

Or Is that a clean install on a different drive

[fierce hollow]

I moved the entire SD folder before cleaning windows. Then (today) I installed git and python 3.10, but as you can see - I can't run the SD

[ornate elk]

Try to run the command it tells you

It can can happen that you have to do it Multiple times for different paths

Better would be a clean reinstall and then only moving models, and outputs over

[fierce hollow]

Do you suggest running the SD again and waiting for it to output on the command line?

[fierce hollow]

(

[ruby yoke]

can you copy the non-english text and paste it here

okay

[fierce hollow]

"To continue, press any key" After pressing it, everything closes

[ruby yoke]

start over and do what @ornate elk told you. remove everything and then do a fresh install. you've got some real confusiong going on there

just save your models and outputs somewhere else and once you've reinstalled, put them into the right directories

[fierce hollow]

Could there be confusion due to changing the local disk? I had SD on disk C, now on disk E

[ruby yoke]

did you read the part of the error i copyed and pasted? this

just uninstall everything and then do a fresh install

[fierce hollow]

Just a second, I'll translate and read it. I'm sorry, I'm communicating with you through a translator.

[ruby yoke]

what language is eaiser for you?

[fierce hollow]

My native language is Russian, but since everyone here is English-speaking, I will write in a language that is convenient for you

[fierce hollow]

Oh... I've already forgotten what and how to install, to be honest, I understand that now I just need to install automatic1111 and then transfer all the models there, lora?

[ruby yoke]

i use deepl to translate

[fierce hollow]

Okey

[ruby yoke]

ошибка говорит о том, что существует путаница в том, кто владеет различными активами - вероятно, потому что вы перенесли программу с одного диска на другой. на данный момент лучшее, что вы можете сделать, это удалить все, переустановить, а затем скопировать ваши модели и выходные данные в нужные каталоги

[fierce hollow]

I understand you, thanks for the detailed translation, now I translate with the help of DeepL, now then I will try to transfer again to the C SD disk, if it does not help then I will reinstall as you said, can you give me an actual link to download automatic111?

[ornate elk]

Yes moving the webui can cause this issue

Also reinstalling windows cause it too. Has sometimes to do with user IDs

[ruby yoke]

не переводите ничего.

1. удалить автоматическую программу 1111
2. установите автоматический 1111 новый
3. скопируйте свои модели из резервной папки в нужную папку для новой установки

[fierce hollow]

Хорошо, вы можете дать актуальную ссылку на автоматическую программу 1111?

[ruby yoke]

https://github.com/AUTOMATIC1111

[fierce hollow]

Thank you kind man!

[ruby yoke]

вы можете рассмотреть возможность установки ComfyUI вместо этого

[ruby yoke]

всегда пожалуйста

[ornate elk]

In the pinned messages of this channel you also find install guides for Auto1111 and the right launch args for your vram amount

[fierce hollow]

Я подключаю плагин в Automatic1111 - ComfyUI, но использовать вместо Automatic я не хочу

[ruby yoke]

хорошо 🙂

[fierce hollow]

@ornate elk Your advice about reverse transfer helped)

Thank you, you are the best 🙂

[ornate elk]

Nice np 🙂

[ornate elk]

You need to add --xformers to the webui-user.bat for best performance

[fierce hollow]

I had terrible performance, so I decided to install from 0).
The settings I had in that SD were like this:
@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--autolaunch --xformers --medvram-sdxl

call webui.bat

[fierce hollow]

Could you please tell me where in the settings you can change the design of SD UI

[honest tendon]

Is it possible to run ZLUDA with a seperate HDD drive?

its taking up too much space

[quaint cypress]

hi, guys, what's wrong with web version of stable audio? model 2.0, audio2audio mode

[fair oxide]

can someone help please?
When trying to launch SD 3 medium, I always get this error.
Already tried reinstalling the file.

Im using the normal stable diffusion webui

[wraith ice]

Hi there
I have a question regarding seeds in the legacy text-to-image sdxl API,
when I pass a seed, and set the image_count to 2, I should get two identical images given the seed is fixed. but the response is two different images. why is that? does the api increment the seed when generating more than one image?

[fading pivot]

Can anyone tell me how to increase the Sampling Steps limit in the webui?
I tried changing "txt2img/Sampling steps/maximum": within ui-config.json, but since I updated the webui to 1.10 it doesn work.
Weirdly enough it does work for "txt2img/Hires steps/maximum":

[thorny iron]

I reinstalled Windows and forgot how to fix this :<

[vocal burrow]

Sd3 is not compatible with auto1111 main branch yet

[vocal burrow]

Can't talk about how the official Api work internally. But yes iutput should be deterministic.
Try with seed 1234 and image count 2. The second image will probably correspond to seed 1235. Easy to test.

[vocal burrow]

Why do you want to increase it? 150 is already stupidly high.

[vocal burrow]

Two things:
1/ you didn't install the client correctly / using "git clone"
2/ delete the venv folder that refers to your previous python install.

To fix your git situation reinstall following pinned guides or run this in your sdwebui folder

git init
git remote add origin https://github.com/AUTOMATIC1111/stable-diffusion-webui.git
git fetch --all
git reset --hard origin/master
git branch --set-upstream-to=origin/master master
git pull

[azure magnet]

I''ve add precision full to the webui but now it give me this error, should i remove it?

[vocal burrow]

What s in your webui-user.bat ?
What s your gpu ?
Screenshot your whole webpage so we can see the settings you just used

[azure magnet]

3060 12 gb

[fossil tendon]

Hi guys, I see this when launching comfyUI :

Device: cuda:0 NVIDIA GeForce RTX 3090 : cudaMallocAsync

Should I do something like installing cuda manually? I use the latest portable version of ComfyUI

[vocal burrow]

What about the webpage?

[vocal burrow]

No you re good

[azure magnet]

Do you mean this?

PS Do i need to upgrade something? this is my version: torch: 2.1.2+cu121  •  xformers: 0.0.23.post1  •  gradio: 3.41.2

[azure magnet]

Fixed

thank anyway

[gentle lava]

can someone help me?

[ornate elk]

Yes its possible

You also can move only the models and add the path to the webui-user.bat with
--models-dir "D:\examplePath"

[ornate elk]

Remove --precision full and dont use --no-half

[azure magnet]

Solved i remove --precision full, Do i need to upgrade something? this is my version: torch: 2.1.2+cu121 xformers: 0.0.23.post1 gradio: 3.41.2

[ornate elk]

No thats good

[fair oxide]

Wdym? How can I run the model, then? Or when will it work on the SD webui?=
Btw, I got the issue fixed (turns out the auto update didnt work), but now have an entirely new problem of,
"log_vml_cpu" not implemented for 'Half'. and when adding the command lines "no-half, full precision" I get another error on every mdoel:
"AttributeError: 'MMDiT' object has no attribute 'model_channels'"

[ruby yoke]

чтобы изменить внешний вид интерфейса, необходимо установить тему comfyUI.

[ornate elk]

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

[misty marten]

The first time it opened normally, I restarted the PC to install ROCm and it doesn't open anymore

[ornate elk]

Have you installed hip SDK 5.7.1 or 6.1 ?

[misty marten]

Yes, 6.1

[ornate elk]

Make sure your GPU driver is updated

Did you also replaced the Library files of rocm?

[vocal burrow]

Nevermind auto1111 main branch got updated 12 hours ago. Sd3 should now be supported

[wild lodge]

can we ask lora training quesitons here?

[misty marten]

yes, I made the old folder and added the new one

I will reinstall everything

[feral otter]

I just updated to the latest Forge UI version and now all my Pony models won't work right. Everything comes out as a non-descript blog. I am using the exact same settings I had used prior to the update and they worked fine. I have all the "score_9, score_8_up etc..." in the prompts and I'm using clip skip 2 but the images are just blobs of color.

anyone have any idea what may have caused this?

[ruby yoke]

someone else reported issues with the forge update they did last night. go open an issue on the github page

[feral otter]

will do

[misty marten]

I reinstalled and now this appeared

[ornate elk]

It says commithash none?
Did you used git clone?

[misty marten]

yes, I did as in the step by step, I created a folder on disk C and inside the folder I opened the cmd through explorer and copied the git clone command

[ornate elk]

Thats strange. Because then it wouldn't show no commit hash

There is something in a language I can't read

Saying something about git error

[misty marten]

this is saying, "git" is not recognized as an internal or external command, operable program or batch file.

[valid plank]

Python got a new version, 3.12.4 does that work or should i still use Python 3.10.11?

[ruby yoke]

do you not have git installed on your machine, and in your system path?

[misty marten]

I have it, but cmd is not recognizing it, I will reinstall it

[ruby yoke]

you can go up to 3.11 - i wouldn't go to 3.12

[ruby yoke]

put it in your system path

[valid plank]

thx

[ornate elk]

Keep using 3.10.11

For best compatibility

[misty marten]

Now I opened the web-ui, but when I click to generate an image, I get this error

[ruby yoke]

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

set your machine to not use the CPU

[misty marten]

and where can i change this?

This also appeared before opening the UI

[ornate elk]

That means you have to do the rocm library steps again

[cold quiver]

If I keep generating will the gb increase? And does it work on a laptop?

Please

[ornate elk]

Which GB?

RAM, vram, HDD space?

And yes stable diffusion works on laptops if they have a good GPU

[ornate elk]

try removing the 1.5 ema pruned model and download dreamshaper v8 to test

[true tinsel]

Hello, today I launched web-ui and gave an error that the versions of pip and pytorch_lightning were not matched. I tried to solve the problem using chatGPT which recommended installing an earlier version of pip but it didn't help, I did it several times, along with rebooting, but the error is still there

[lunar perch]

Hi could you please advise me on how to format/partition? I got a new 4tb SSD I want to install Linux and automatuc1111 on it and use it to store all my big files (models, loras, etc), AND use it sometimes for webuis in windows as well (SD.next, forge). So I want to reuse the same directory for loras and models across windows and Linux dual boot, issue is, sharing requires ntfs and I understand having the models on an ntfs partition will slow down the Linux perf of automatuc1111? Is that correct?

The whole reason I'm going this is bc I have AmdGpu and SD is supposed to be way faster on Linux for AMDGPU

[true tinsel]

I also tried to resolve dependency conflicts, but I was able to figure it out, does anyone have any thoughts?

Update:ok, I realized that the problem is in the pydentic version, it needs to be changed to 2.2.0, for my reasons, it will fit the conditions of fastapi and gradio. I tried changing the version of pydentic, installed it and changed the values ​​in the requirements.txt file but it still showed version 1.10.15. How can I change the version of pydentic I'm using?

[somber oriole]

Hey I'm using a 1070 8gb with webui forge, was wondering what kinda models and loras would be stable to use with a 1070 8gb

[ruby loom]

HI

[ornate elk]

Hey, try this:
Install python 3.10.11 64bit
Then delete the venv folder and relaunch the webui-user.bat

[ornate elk]

Hey, 1.5 and sdxl/pony will work

[ornate elk]

Hey what's your GPU?

With zluda, you can have nearly the same performance as on Linux.
So better use that.
I made a lot of AMD guides for windows. With zluda or directml. Check out the link in the pinned messages.

[oblique vapor]

Question, how long generally would one expect an image to take to generate? I have 32 gigs of ram and an 8 core processor. I have 30 steps in a 52 token image. Sometimes images are really fast, other times it feels like it takes quite a while

I'd like to speed up image generation if possible 🤔

Ideally just general performance improvement

[lunar perch]

@ornate elk amd Rx 7900xtx 24g ram

Yeah now I believe win finally has rocm drivers rught

But my understanding is that it's still faster on Linux and main benefit us u don't need some weird fork like AmdGpu Forge or SD.next

U can use the standard automatuc1111 like everyone else so u can use all native features

No need to stumble over extra issues of the alternate amd version

@ornate elk I already use AmdGpu Forge on windows and it's pretty fast

But I just want the optimal setup

Idk people told me the windows version has major bottlenecks like can't actually use all the VRAM of the gpu

[delicate citrus]

is colab not working for sd? I have problem to run a11 and comfy

[ruby yoke]

what issue are you having?

[delicate citrus]

@ruby yoke

[ruby yoke]

read the last line of the first screen shot

[delicate citrus]

for a11 I dont understand, for comfy looks that was my drive, I'm starting again

🆗 comfy looks ok, I think was my drive

[ruby yoke]

it says "ImportError: cannot import name '_marching_cubes_classic_cy' from partially initialized module 'skimage.measure' (most likely due to a curcular import)
/usr/local/lib/python3.10/dist-packages/skimage/measure/init.py"

so it looks to me like it's haivng a problem with python

[delicate citrus]

@ruby yoke hmmm ok, I will try something with chatgpt, thanks

[delicate citrus]

what is this on comfy? "idle"

[ruby yoke]

that's a comfy with no workflow just sitting there. load a workflow, load the default, or at least load some nodes and do something

[unique field]

Had a working environment but I nuked it to reinstall comfyui and the manager via the terminal

got everything reinstalled and I'm coming across two things, when I try to run the program, my error log says it's missing dependencies at this point. Is this a case of my python install being too old?

My current version of pip is 22.2.1 vs the maximum of 24.1.2

but I do recall that it was not advised in automatic to get the latest pip ver. Is that still the case here?

that said, if the current version works best then I can update that really easy.

Or should I try to get a specific version of python for the most optimal experience?

[ruby yoke]

my python is 3.11

[unique field]

Hrm, just did a new install and updated and restarted the comfyui manager through custom nodes manager here. But when I try to hit the "update all" or "update comfyUI" it gives an error and I'm not sure what I'm doing wrong at this moment. Fetch Updates simplays says "Everything's up to date"

[oblique vapor]

Question: I have noticed that the GUI seems to be missing certain Loras. I think it's cool to have all of my Loras displayed in the Lora tab, but not all of them are there. I have closed everything down and re-opened it, but some simply refuse to appear. Anyways I can fix that?

[ruby yoke]

you are probably going to ahve to uninstall everything, make sure your python's at the right level, then do a fresh install

[ruby yoke]

make sure that they are in the folder they are supposed to be in, and that they aren't partial files that are still in the middle of downloading or something

[unique field]

alright, I should just nuke my current stuff then.

[fair oxide]

I suggest git cloning the repo instead of messing with portable installer

easy way to install comfyui on windows manually

[unique field]

yeah, I ran the following commands, python -m venv venv venv\Scripts\activate pip install comfy-cli comfy install

said commnd was the "Recommended way" to go about it, but it did not include the runtimes

[fair oxide]

comfy-cli, are you on linux?

[unique field]

this particular install is on windows, my linux server is running it fine

[fair oxide]

i see

you can have a look at the guide i just shared for windows

[unique field]

I'll take a look at that, thanks

though, I may need to nuke my old installtions

[fair oxide]

should be a walk in the park, good luck

[unique field]

before I do anything else

regardless, thank you for that link

[fair oxide]

np, and yes i would also suggest doing a clean install

[unique field]

Yeah, not exactly the most used to windows tbh

not as easy to nuke previous versions

[fair oxide]

feel free to ask for help in case you find any step confusing

[unique field]

well, do I need to wipe out my current python install?

[fair oxide]

not really, if you have python and git installed already, you could just remove the comfyui folder and install just that

back up any necessary files before wiping

like checkpoints, lora

[unique field]

All set on that front.... this is the first time I'm using comfyui, all the checkpoints and loras are in my automatic folder

[fair oxide]

ahh ok

no harm in keeping a1111 along side comfyui if you have decent storage

[unique field]

the windows machine just has a lot stronger gpu

"laughs in high speed nas"

but in terms of acutal ssd storage I have on this machine.... 10 TB

so, not exactly an issue

[fair oxide]

great, and given your net is also good it shouldnt take more than 10 to 15 mins max to install comfyui on windows

[unique field]

yup,

that said, I'll be double checking to make sure the files extract properly

I'm curious how my Xeon cpu's will handle this comfyui tbh

prob not very well due to them being old

[ruby yoke]

what CPU do you have, and what GPU?

[unique field]

for the windows machine I'm using?

[ruby yoke]

and for that matter, what version of windows?

[unique field]

3090 nvidia, threadripper 3990x. Got the cpu for free from work

windows 11, fully up to date

[ruby yoke]

intel, amd, what?

[unique field]

amd threadripper 3990x

[ruby yoke]

ummm - you want to make sure it doesn't try to do the graphics processing

[unique field]

Oh I know

[ruby yoke]

then you should be happy with the performance

[unique field]

Oh yeah, performance is not the issue

256 gb ram as well.

[ruby yoke]

you said >I'm curious how my Xeon cpu's will handle this comfyui tbh< there's not a lot for the CPU to do, really, i'ts all on the GPU, so it should handle it fine

[unique field]

Yeah, I have a few poweredge servers

with those

no gpu and its more of a curiosity if they would even manage to run ai art at all

morbid curiosity

[ruby yoke]

it'll run, it'll just run like frozen molassas in winter

[unique field]

That's what I had figured. lmao

[ruby yoke]

unless they can't run cuda, then maybe about the speed of glaciers

[unique field]

the threadripper MIGHT do an OK job

just because this thing has so many threads

[ruby yoke]

guess it depends on if it can actually run the basic stuff - like python

[unique field]

It can run crysis by itself without a gpu.... so yeah

[ruby yoke]

would be an interesting experiement

[unique field]

Yeah, I can let you know once I get everything set up and see how bad it performs

because I'm not expecting much from it

gpu > cpu every time in tasks like this

[ruby yoke]

yes, well - there's a reason that the world is currently running on nvidia

[unique field]

Nvidia and ARM

[ruby yoke]

?

[unique field]

ARM is extremely power efficient machines that handle all sorts of servers and machines

ARM is honestly incredible in how low power usage it is

[ruby yoke]

they're not backordered though, everyone's ordering nvidia

[unique field]

different use cases

[oblique vapor]

After doing some digging and looking, I think it could have something to do with the base model. It appears that those that show up have a base model of SD 1.5. But here's something really strange, if I put a different Model Lora, like a Pony Lora or a SDXL 1.0 Lora in the checkpoints (There because I was just just starting out and I had a lot to learn) It not only shows up in the Lora tab, but ALSO lets me run without giving me a blob.

[ruby yoke]

oh. yes. you can only use loras with the model they were created for. so if you have an SDXL model loaded, you're only going to see the SDXL loras, etc

[oblique vapor]

Very interesting. Then how come I can see them in the checkpoints tab?

And if that's the case then should I be doing something in order to make certain Loras more accessible? This raises a lot of questions

[ruby yoke]

they're on your machine? but if you try to use a lora that's written for a different model - a 1.5 lora with an sdxl checkpoint for example - you'll have a mess - the weights ren't the same in the base models, so the updated weights in the lora won't work

[naive badger]

I can't use extras to upscale and get this when i try to generate. [WinError 123] The filename, directory name, or volume label syntax is incorrect: 'outputs\extras-images\00366.png?1722136620.153402'

[oblique vapor]

I phrased it poorly. What I mean is, if I put a Lora that is meant for a different model, say for example a Pony Lora in the SD folder like I would a check point, it presents itself like it's a check point. Does this sound like it should be happening?

[naive badger]

it is saving it but not displaying it in webui

[unique field]

Regarding ComfyUI's Models/Checkpoints directory, can you put sub directories in there such as 1.5, SDXL, and so on?

or do they have to be in that very specific directory in the checkpoints?

[ruby yoke]

you have to put the loras in the lora folder.

[oblique vapor]

Right

And what I am saying is that, out of ignorance, I did not do that. And they appear as check points.

[ruby yoke]

move them to the right folder, restart, they should show up correctly then

they don't work as check points even though they show up that way. not even information in them

[oblique vapor]

Makes sense

This may be a dumb question but, is there a way to change the base model?

[ruby yoke]

a check point updates a LOT of the base model's weights. a lora updates just a very few specific weights.

[oblique vapor]

Or would that happen just via changing the checkpoints?

[oblique vapor]

I see

[ruby yoke]

nope. think of it this way. you have a game. the game is your base model. you have house rules for the game, so when you play at home, you ahve a small sheet with the house rules on it. they overwrite, for that game, how you play the game. the house rules sheet is the lora.

[unique field]

would this work?

[ruby yoke]

you cant' use those rules with any other game

[ruby yoke]

yes but you'll ahve to remember that you've got your sdxl check points in the SDXL folder. and other people that just put all their checkpoints in one folder won't find yours easily.

[oblique vapor]

I see. Tyvm btw for all of this information! I really appreciate you helping me with all of this.

[ruby yoke]

you're welcome. sorry for the confusion

[oblique vapor]

So then what do I do if I find a SDXL model or Lora I wish to use?
And no worries you're good! Learning something new isn't easy and I am here for it. I just appreciate your patience. :)

[unique field]

Yeah, I just have them divided up by sdxl and 1.5

[ruby yoke]

you download it, and put it in the right folder, and then restart the interface.

[unique field]

and then sub folders/directories from that

such as, landscape, anime, realistic, and so on

in each one

[ruby yoke]

my checkpoints looks like this

however don't make too many sub folders - you'll wind up with a path that's too long

somethign like: checkpoints -> folder -> folder -> folder don't do that

[oblique vapor]

So you can make sub folders? Very interesting.

[unique field]

Yeah, too long of a path is not good

its just stuff like, SDXL -> Anime, Realistic, and so on

and that's the last level

[ruby yoke]

yes. but here's the issue - it'll work for you. share your workflow with someone else and they're not going to have the model in the folder path you do, so they'll have to go find it. also, if you load someone else's workflow, chances are good that you'll ahve missing models - even if you have them - as your folders are not standard.

and only do that for checkpoints. everythign else, put exctly what it has to go

[unique field]

Yeah, I made that mistake in automatic, forced myself to do much better on the organization side of things after that

because what's the point of having said items if you cant find them?

all the safetensor files are behind that layer

Hrm, followed that vid to the letter and got fbgemm.dll missing.

going to reboot and see if it's just not a nonsense on that

Hrm, one reboot later and still getting that error

it seems I'm missing the Nvidia start.bat file again

nuking my python install and trying that again just to make sure it's not that

[oblique vapor]

Understood, but follow up question. I just spent a while sorting everything and so now everything should probably look a bit like what you have. (if significantly smaller)

Though, you're calling it a workflow. I was under the impression workflows were automated processes. Am I misunderstanding?

Side note, going to run some tests, this could be really cool if I've hooked it all up correctly

[ruby yoke]

the workflow are specifically the nodes and how they are connected inside comfy- not your directory structure

[oblique vapor]

Hmm okay... Question, can you tell me the basic components of a simple workflow?

I've seen one could go download a bunch of workflows off Civitai. What might I be able to expect if I downloaded one?

[ruby yoke]

sure. very basic is - you need something to condition with - either a text prompt or an image. you need a model of some sort. you condition the model's response with the prompt (image or text), you run it through some sampling math, and it generates an image.

here - this is a simple SDXL workflow

download it and upload it into comfy - either with the load command or just drag and drop it into the interface once it's running

[unique field]

trying to install an older version of torch torchvision and torchaudio and see if that helps with anything.

[ruby yoke]

VENV

[unique field]

yeah, venv failed

[ruby yoke]

ouch

[late tree]

Anyone know what to do about this? When I start up SD it shows an error for ADetailer. This is what it shows:

AttributeError: 'OptionInfo' object has no attribute 'needs_reload_ui'

Both the UI and ADetailer are up to date.

[oblique vapor]

I think I am beginning to understand! So a workflow is a lot like a process? It sounds like it's something one creates every time they make a new image, is that correct?

[unique field]

(bangs head)

Torch not compiled with Cuda enabled

[ruby yoke]

my current comfy interface and the nodes on it are my current workflow

it's how the data flows from one node to the other (you are programming using a graphical interface)

[ruby yoke]

dude - go eat, go take a walk, go do something mindless for a while. recharge your brain.

[unique field]

just ate a bit ago, haha

and tbh, I'm stubborn

[ruby yoke]

unplug. seriously. or go make music with suno for half an hour

let your subconcious work on it without your concious interferance

[oblique vapor]

Ahh and this is comfy UI? Currently I believe I using Auto1111, I might need to download some extra parts to be able to see the same stuff.

[ruby yoke]

this is comfy, but the principal is the same for auto1111 - what you do, and how you ahve things called is a workflow

[oblique vapor]

I agree friend, a good break works wonders on a programmers brain. Take a shower and destress and it will be easier to make forward progress.

[ruby yoke]

you can have both installed, and just share the models between them

[oblique vapor]

I think I gotcha

[ruby yoke]

@unique field flashes of insight only happen when you're not thinking about the problem

[inner lodge]

So I use Forge through Stability Matrix and saw there was an update. I tried to download it, it said update failed, and trying to launch it gives me this. Looking on the github, someone else has posted this same issue and others are making mentions that the update broke forge. Not sure if there is anything I can do on my side or if this is one of those cases where I just have to sit and wait until it is fixed, I am not too in the know about python or gradio, so most of the data listed is lost on me.

[ruby yoke]

seveal people have reported serious problems from that update. go to the github page and open an issue

[inner lodge]

Seeing as someone else has already done so, it feels like it might be unneccessary to put the same thing in again? I guess I could follow up on the issue and make mention that I am having the same issue. Either way, seems like I might just need to sit and wait which is annoying seeing that I only got back into doing stuff two days ago

[ruby yoke]

it's necessary. the more reports they get, with screenshots, the easier it will be for them to pinpoint what happened

[inner lodge]

Alright, I will look into it then

[ruby yoke]

just be as detailed as possible

[rich garden]

I tried adding some embedding files into the folder but they didn’t appear in the Stable Diffusion browser under the “textual inversion” tab also what does the embedding stuff do anyway? And while I’m at it, what do VAE’s do?

PING ME

[unique field]

so I figured out the problem

the current pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu121 install of this command on the webpage will just fail every time

it will always result in the cuda error

using the other nvidia command for Nightly

pip install --pre torch torchvision torchaudio --index-url https://download.pytorch.org/whl/nightly/cu124

worked the first time

the torch versions that worked are are follows,

Torch 2.5.0 Torchaudio 2.4.0 Torchvision 0.20.0

the one that DID NOT work was torch 2.4.0 TorchAudio 2.4.0 Torchvision 0.19.0

and this is on 22.2.1's version of pip

everything works as intended now. Phew.

so if anyone else is getting the same nonsense I was earlier, try the nightly version

[fathom rose]

can somebody help me? I could not install xformers in latest version auto1111 using the webui-user commandlines. I'm trying to manually put the folders in the right place. I need the latest version of xformers in their associated folders. Can you please share with me?

is this the one currently in use or there is any update?

[dense tapir]

Hey guys! How do I change my default settings like prompts and basic generational settings? I did it before but simply forgot which file to edit.

[unique field]

which ui are you using?

[fathom rose]

how do i fix this?

[oblique vapor]

Random quetion before I lose concsiousness (Kinda sleepy)
Embeddings aren't restricted to a model correct?

Like, if I have "verybadimagenerative" as an embedding, I can use that regardless of the checkpoint?

[ruby yoke]

i think i'm going to let you sleep on that and do some research on embeddings when you wake up

[safe rover]

When you said this, did you mean using the InstantStyle extension?

[fair oxide]

bumped into that cuda issue too, fixed it, here is how

[thorny iron]

is it borked

[ornate elk]

Hey, I have the same GPU.
In using it with Auto1111 fork with Zluda. It has the best speed/usage for me.
There is now WSL2 on windows. With that you can install Auto1111 in a Linux VM which has the same speed as native Linux because it uses the new Rocm 6.1 drivers.
But Linux and wsl have both a big problem when it comes to memory management (RAM).
They use near Max of it the whole time while zluda has a better optimisation.

[ornate elk]

Commit hash none is not good

Seems like you didnt git cloned it

[thorny iron]

I guess I will delete and start again.. :<

[ornate elk]

Checkout the nvidia guide in the pinned messages

[ornate elk]

Its an out of vram error. So either adjust some settings or when it happened while upscaling enabled the tiled vae extension

[ornate elk]

Embeddings also need a compatible checkpoint. Means 1.5 for 1.5 and sdxl for sdxl

[ornate elk]

Same as for embeddings. Loras will only show up when a compatible model is selected

[fierce hollow]

Is there a Lora that can create documents ? (Passport or other ID card) I am interested in this for purely introductory purposes, don't ban me)

[fierce hollow]

Is there a guide to Stable Video 4D?

[vernal sundial]

new to animatediff. just a question why are outputs that use control net so blury? this is the results i get using canny on some stock footage of a buff guy doing curls - with superman in the prompt

its just 32 frames at 25 fps

and yes i know its a gif with colour clipping, but even on the mp4 its still super blurry

also he looks like his in a rave but i can assure you the stock footage is not set in one lol

[ivory bramble]

Ok, I'm starting over again, 4th time with the NVIDIA guide. I installed Python 3.10.11. web-user.bat has set VENV_DIR= set COMMANDLINE_ARGS= --xformers --no-half-vae --medvram and now it's opened on browser. So far so good. Besides ControlNet, what are the essentials I need/should have?

The previous installation was shutting down my PC and when it wasn't, it gave me strange results. I don't know if it had to do with re-naming models, but I did that so I've deleted all of them and I'm going to start fresh. What is the best model for photo realism in your opinion?

What are some settings I should change in Stable Diffusion? There was a word that many model had "baked in". I have a horrible memory and I'd search if I knew the name of it but I do remember changing the setting with it.

VAE

https://i.imgur.com/8iPVGzE.png

Which one do I choose again? Sorry for the flooding of messages. I'll keep it toned down.

[wind roost]

im in a bit of a predicament guys

idk what setting i have messed up but my renders are always blurry

here are my settings

im so confused and its really irritating me if anyone knows what I am doing wrong please ping me

how do i increase the sharpness

for information i am on an RTX 3070 on driver version 536.23

using comfyUI, get same issue on automatic1111

[somber oriole]

Anyone else having problems searching for models on civitai? I keep on using the search bar and it's just stuck on searching over and over

[ivory bramble]

After fresh install, didn't change anything, typed 'cat' in txt2img, PC shut down again

I had zero issues with this in my first installation

I don't get it

[wind roost]

how many iterations should i do and what cfg scale with dpm++3m , it always seems to have issues, do i need to alter nvidia control panel

is it the denoiser?

[pastel sequoia]

Hello, first of all, I'm on AMD and I'm trying to use SD-CN-ANIMATION and I'm getting this error, what should I do?

[ornate elk]

You need to increase your windows pagefile to not get crashes

Thats important because you only have 16gb ram, but sdxl models need more

[ornate elk]

You need to run
Pip install basicsr
Inside the venv in a cmd

[fierce hollow]

Is there a guide to Stable Video 4D?

[steep cape]

I need some help with LoRA training, I just tried my first training in Kohya after following a guide. I set it to 10 epoch, expecting it to create 10 different .safetensor files, but the training ended after creating only 2 files. Should it not keep going to 10? I attached screenshots of my training parameters

I'm trying to train a realistic, custom made character into a LoRA if that helps

Also, the files that it generated are 1.6GB in size, thats not normal, right?

[hallow latch]

Is there a way to removed certain Upscalers like you can do with Sampling methods?

[fair oxide]

Hey, I have an NVidia 3060ti. THis is in mz webui-user.bat:

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --disable-safe-unpickle --autolaunch --medvram --api --port 30000 --cors-allow-origins=https://www.painthua.com --xformers --medvram-sdxl --no-half-vae
git pull
call webui.bat

[unique field]

Trying to follow a tutorial but I cant seem to find this node anywhere, is this deprecated?

[karmic crown]

what manager returns if you search for controlnet?

i would probably just use this
https://github.com/Fannovel16/comfyui_controlnet_aux

[unique field]

Yeah, that's the one I was looking at and installed

[karmic crown]

see if this can help
https://medium.com/@promptingpixels/how-to-add-controlnets-to-comfyui-960e54106d29

[unique field]

yeah, that more or less answers that question that the controlnet preprocessor there is deprecated

[ornate elk]

Remove --medvram --disable-safe-unpickle

[unique field]

I'm unable to change the undefined to anything

What do I need to do here?

am I missing models?

[karmic crown]

did you download contronet models?

[unique field]

I had not, I just did now.

Any must haves?

[karmic crown]

i have all (somewhere in backup) 🙂

https://huggingface.co/xinsir/controlnet-union-sdxl-1.0

https://huggingface.co/lllyasviel/sd_control_collection/tree/main

[true tinsel]

@inner lodge
Hi dude, I found a solution to the problem. You need to find a file called "requirements_versions.txt" in the folder with Stable Diffusion And there change the value pydantic==1.10.15 to pydantic==2.1.1. Save and then run again

[inner lodge]

I saw a similar thing on the issue page, but what I ended up doing was deleting pydantic==1.10.15 and that fixed it for me

not sure if it will have consequences later or not

[true tinsel]

ХD you are not looking for difficult ways) In general, I solved my problem, if you want, you can use the solution

[inner lodge]

Yeah yeah, either way thanks for the help

[karmic crown]

nope, it is fine
it is installed automatically by gradio anyway

btw. pydantic was removed from requiremetns_version.txt in the latest version on github

[inner lodge]

ah ok, so probably will update it here in a bit then

[fierce hollow]

Is there a guide to Stable Video 4D?

[sage bramble]

Hello all. I just started using SD, trying to pair it up with Ollama and Open-webui. I had it working yesterday with no issues, until I made a mistake and had to redo my entire Ubuntu drive. Anyways, I've followed all the commands, I made a new conda env, with 3.11.9 Python installed, and I try to run SD and I'm getting this error:

~/stable-diffusion-webui$ ./webui.sh
Install script for stable-diffusion + Web UI
Tested on Debian 11 (Bullseye), Fedora 34+ and openSUSE Leap 15.4 or newer.
Running on jmp470 user
Repo already cloned, using it as install directory
Create and activate python venv
ERROR: Cannot activate python venv, aborting...

any tips on where to start troubleshooting?

[ruby yoke]

is that the extent of the error logs? i think you're going to be recreating your venv

[sage bramble]

yeah, that's what I'm afraid of 🙂

I created an env called LLM, for open-webui. It needed python 3.11.x, as my base python env has 3.12.x

[ruby yoke]

that's what your error says, it can't activate your environment. start by trouble shooting your venv

[unique field]

is there a way to make positive and negative prompts be imported via another program for ComfyUI?

[fair oxide]

Okay, but what about my original question 😅

[ornate elk]

Would need a full cmd log

If you have any error

SD3 works in the latest auto1111 version

[oblique vapor]

Thank you very much!!

[lean glade]

So I've been trying to get my Stable Diffusion to work and CMD keeps spitting this at me, tried uninstalling and reinstalling onxxruntime to no avail.

Any advice would be appreciated.

Also tried running it through the venv and get the same error.

[worthy sapphire]

So I have been trying to get SD to run on my amd gpu through zluda, but when I try to generate through the webui I get this error message. "RuntimeError: CUDA error: invalid argument CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect. For debugging consider passing CUDA_LAUNCH_BLOCKING=1. Compile with TORCH_USE_CUDA_DSA to enable device-side assertions." I thought I followed all the steps from the guide I was going off of.

[quick arrow]

No.

They're not out yet.

[wintry atlas]

I am on the latest Auto1111, trying SD3 and keep getting this kind of output

any idea why?

Steps: 30, Sampler: DPM++ 2M, Schedule type: Karras, CFG scale: 4, Seed: 3463910682, Size: 1024x1024, Model hash: cc236278d2, Model: sd3_medium, Hypertile U-Net: True, Hypertile U-Net max tile size: 192, Hypertile VAE: True, Hires CFG scale: 3.5, Downcast alphas_cumprod: True, Discard penultimate sigma: True, SGM noise multiplier: True, Version: v1.10.1

[ruby yoke]

SD 3 doesn't work with karras

[wintry atlas]

ok, it's actually set to automatic

[ruby yoke]

set it to simple and use huenpp

also, SD3 does NOT use U-net

so get rid of anything that talks about unet

Hypertile U-Net: True, Hypertile U-Net max tile size: 192< you can't use that with SD3

[wintry atlas]

it be nice to have a warning or something

[ruby yoke]

on the page where you can download the sd3 safetensors files, there are sample workflows

[wintry atlas]

for auto1111?

[ruby yoke]

no. Auto1111 have very little function for SD3, though there is some functionality. not sure if you'll find workflows on the auto1111 github page or not

[wintry atlas]

right, which is what i am using

[ruby yoke]

yes. and -no, on the huggingface page there isn't anything for auto1111, you'll need to look on the auto1111 github page to see if there might be something there

[wintry atlas]

does the --precision command line param have any negative effect on sd3?

[ruby yoke]

i'm not sure, i use comfy - so i'm not sure what that flag actually does for auto1111

@proper swift can you answer the --precision question?

[wintry atlas]

well, i disabled hypertile, set the sampler to heun, type simple

"improvement"

[ruby yoke]

yes. hang on and let's see if icey responds

[wintry atlas]

ill disable some other things

[ruby yoke]

if not, @ornate elk might

[wintry atlas]

Downcast alphas_cumprod: True, Discard penultimate sigma: True, SGM noise multiplier: True

it used to work before but only in 512x512

but that was like a month ago

bingo

[ruby yoke]

very nice 🙂

you might save your configuration

[wintry atlas]

im going to see which setting did that and report it

[proper swift]

i haven't tested any precision changing for any ui. i'm using default sd3 all the way and have only been exploring the prompting with it so far

[wintry atlas]

does upscaling work with SD3?

[ruby yoke]

you mean during generation or after the image is finished?

[wintry atlas]

after

[ruby yoke]

once the image is finished, it's an image. run it into any upscaler you want

[wintry atlas]

ok, then hires fix, which i am not sure exists in comfi

[proper swift]

i think if you use sd3 for upscaling, you have to use tiling. it doesn't like going out of the megapixel resolution range i've noticed

[ruby yoke]

hires fix is just the different gans and some tweaking.

[wintry atlas]

well it's giving me an error

RuntimeError: The size of tensor a (115) must match the size of tensor b (114) at non-singleton dimension 2

[proper swift]

don't think hires fix works with sd3 but i haven't tested that yet. i've tried generating direct to 1500x1500 and it fails

[ruby yoke]

don't upscale while you're generating, just create the image and then upscale it if you realy need it larger

no point in wasting compute time making a huge image you don't really need to be huge

[proper swift]

hires fix is img2img at the core. you can use a gan upscaler as a prescaler before the denoising

[ornate elk]

What's your GPU?

[ornate elk]

Can you show the me full cmd log?
Also did you followed my Zluda guide?

[ornate elk]

Try Euler a as sampler

[lost crescent]

ComfyUI Xformers says it requires torch 2.0.1+cu118 - but other nodes demand torch 2.3.1+cu118. Why the conflict?

[lean glade]

AMD Radeon 7700 XT

Been at work, so sorry for late reply. Lol

[fathom rose]

what is this error about? i'm using forge ui and sd3

here's the whole thing

[ornate elk]

With that GPU you should follow my Automatic1111 Zluda AMD guide.
It has a much better performance than directml.
You find the latest guide for that in the pinned messages. There at the top link to my Github.

[ornate elk]

Try using the 5gb SD3 model which had the clip text encoders included or go into Settings and enable T5 text encoder

[lost crescent]

I cannot use xformers because torch2.3.1+cu118 won't work

[fierce hollow]

Is there a guide to Stable Video 4D?

[ornate elk]

You need to upgrade the torch version to a compatible one with the latest torch

[fathom rose]

not working at all

[lost crescent]

The problem is that Xformers asks for torch2.0.0+cu118; and then when I hyave d/loaded and installed that version, during ComfyUI Startup - a custom node demands - and d/loads - torch 2.3.1+cu118 ... overwriting the correct version of torch for Xformers!

So I need to upgrade Xformers to a version which is satisfied by torch2.3.1?

[ornate elk]

Set clip skip to 1

And Euler a as sampler

[fathom rose]

it's showing the same error

[ornate elk]

Can you show a full cmd log?

[floral fable]

i paid for the chat bot but i cant sign in

[lost crescent]

Dang! I upgraded to xformers 0.0.27 - compatible with torch 2.4.0+cu118 - but when I start ComfyUI - torch 2.4.0+cu118 is overwritten by torch 2.3.1+cu118 - and not compatible!!!

[ornate elk]

Try xformers 0.0.25

[ornate elk]

Make sure you logged into the stability ai site and linked your discord account

[gaunt vault]

Hello I am looking at getting a license but not sure if I need community or enterprise license. I am a business running less than $1million a year

[floral fable]

it just clicked login and it just go into loops (and just re direct me to Choose the right plan for you. )

[ornate elk]

Best to write an ticket mail to Stability ai directly

[ornate elk]

Very strange. Try login here:
https://stability.ai/stable-artisan

[lean glade]

@ornate elk Just finished work and got home, up to downloading the AMD HIP SDK 5.7, says to avoid the PRO version but there isn't another option?

[ornate elk]

Download the 6.1.2 hip sdk

And then when installing just don't enable the pro drivers.

[lean glade]

Oh okay, thanks. Was a little confused.

[ornate elk]

NP, yea its confusing that they name it pro driver

Instead of hip SDK

[lean glade]

Also side note, love the pfp.

[ornate elk]

Thanks!

[lean glade]

OP is goated. Started rewatching through a react channel and early One Piece hits different after being up to date. lol

Side note over. (Don't want to derail the channel topic)

[ornate elk]

Yep totally!

[lean glade]

Moment of truth

Wait, I restart after installing the HIP SDK?

or is that only specific to the GPUs mentioned?

If this works ima have to netlix and chill you.

@ornate elk Got this, started an image generation, waiting to see if it fails but don't know what this means. lol

[ornate elk]

Yes a restart is needed.
What's your GPU BTW?

[ornate elk]

Let it load

[lean glade]

I already stopped and restarted my PC. lol

[ornate elk]

Oh ok

[lean glade]

https://tenor.com/view/goofy-trial-crazy-evil-ill-fucking-do-it-again-gif-17111514

[ornate elk]

Yea relaunch the webui-user.bat

[lean glade]

Okay, starting a generation of "dog"

lol

[ornate elk]

The first image will take up to 40 minutes

[lean glade]

But after it'll be quicker?

[ornate elk]

The generate button should be grayed out and the cmd shouldn't show any errors or progress

[ornate elk]

Yea after that one time. It will take seconds

[lean glade]

Yeah, greyed out, no errors so far.

I love that Goofy gif. xD

[ornate elk]

Thats good 🙂 then it should work

[lean glade]

Oh I can see the loading bar thing in the cmd now so seems good.

Nvidia users have it easy, omg. lol

[ornate elk]

True xD

[lean glade]

Imagine have a single click install.

[ornate elk]

For SD webuis that's rare

[lean glade]

Curious what dog it's going to give me, any bets?

[ornate elk]

Maybe a golden retriever on gras

[lean glade]

That... Would be so cliche you might be right. haha

I'll guess... German Sheperd.

[ornate elk]

You can checkout civitai.com for community models meanwhile

[lean glade]

Yeah I know about Civit.ai

[ornate elk]

The base 1.5 model is bad so you don't need it after that first image

[karmic crown]

buy mac, and use my app - then you will not have to imagine that 😂

[lean glade]

I upgraded from an Nvidia 1660TI to my current GPU so was shocked how difficult getting SD up and running was. lol

[lean glade]

I also like to play games

https://tenor.com/view/windows-linux-linux-users-windows-users-windows11-gif-25304807

[karmic crown]

i was joking about buying the mac, but i was serious about one click install on mac (a1111 or forge)

[lean glade]

Yeah I know, haha

[ornate elk]

If you want to run LLMs on AMD its easier.
You can use GPT4all or Ollama or LMstudio

[lean glade]

idk wtf kind of dog this is

[ornate elk]

Melted dog

[lean glade]

I don't think I've SEEN this breed.

[karmic crown]

Windows 11 ARM works faster in VMware on my Mac than many real windows machines (in some aspects)

[ornate elk]

But hey it works ^^

[lean glade]

Yeah, thank fuck. lol

Had a meme video prepared if it didn't work. xD

^

Something a little more complex, am happy.

Thanks @ornate elk, amazing work.

[ornate elk]

No problem, the girl reminds me a bit of Bocchi the Rock

[lean glade]

First I'm hearing of this, watching the solo. lol

Yo she shreddin'

[lean glade]

Had to share this
https://youtu.be/YVVaGBe_DXE

[fair oxide]

hi

I am getting an error in the Installing requirements section, can you help me?

My Python version is 3.10.11
My graphics card in this system is Amd Rx580
set COMMANDLINE_ARGS=COMMANDLINE_ARGS=--lowvram --precision full --no-half --skip-torch-cuda-test
(I took this command from github)

Getting requirements to build wheel did not run successfully. exit code: 1

[ornate elk]

Hey, you probably didnt installed the correct version for AMD cards.

Also the commandline_args that you used are wrong

[fair oxide]

Can I send the code I downloaded?

[fair oxide]

I found this on github as a solution to another problem : torch is not able to use gpu stable

https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Install-and-Run-on-AMD-GPUs

I downloaded it from here

[ornate elk]

this information is outdated

please use my AMD guides

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

your GPU is supported by Zluda so that would be your best option performance wise

So try the Automatic1111 with Zluda guide

[fair oxide]

thanks i'm trying now

[ornate elk]

let me know if you have any questions

[fair oxide]

yes actually there is one, even if I add python to PATH it can't find it, so I added it to the Webui-User Python = section, will this cause a problem? Seriously, even though I added it to the system environment variables, it does not appear. It is the same in the automatic adding option in setup. It only sees that I downloaded it from the Windows Store. That's why I added it to the Python = section. The python I downloaded from Windows causes other errors, it cannot find the setuptools.

does it create a problem?

[ornate elk]

dont download it from the windows store

that is a problem

[fair oxide]

No, I already downloaded it from its own site, but since it could not detect it, I added the file path to Webui user

[ornate elk]

ah okay

thats okay

[fair oxide]

Dont check the Pro Driver. What do you mean here? I didn't see such an option. There are ray tracing libraries and time controllers when installing.

[ornate elk]

at the bottom it normaly says amd pro driver "disabled"

so leave everything as it is and your fine

[boreal iris]

I haven't opened it in quite a while... why am i getting this error when i try to generate in img2img?

[ornate elk]

it says cannot identify image file

make sure its png or jpg

[ornate elk]

try rename it to image

[boreal iris]

I just copied it to paint.net saved it as another png file and it worked

ty

[ornate elk]

Nice np

[smoky mesa]

Hi all, sorry for the bother, i am hoping some one could help me, this might be hard for me to explain but i will try my best.

when you click the CivitAI icon at the top of a checkpoint image within automatic 1111, it brings up the data of the checkpoint, and when you scroll down, you can see images and the data for the image as shown in this picture,

what could be the reason for the info not showing up anymore, as show in this picture?

If any one can assist me, it would be highly appreciated, tried updating extentions, reinstalling SD Automatic 1111, dont know what else to do.

[ornate elk]

People can upload images without providing meta data to the image

[smoky mesa]

that i understand, but in this case, nothing has any meta data, least now shown, where it usually did.

[ornate elk]

What does the cmd shows?

[boreal iris]

Why do my outpaints look so bad though?

[smoky mesa]

im sorry, i dont follow, the cmd?

[ornate elk]

The black window

That runs in the background

[smoky mesa]

When i start it up for the 1st time? should i put it in a text file?

[ornate elk]

Yes

You can copy and paste it in here

[smoky mesa]

alright, give me a second, thank you for your help.

venv "C:\STABLE DIFFUSION\sd.webui\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.
CivitAI Browser+: Aria2 RPC started
Loading weights [f5e48cafaa] from C:\STABLE DIFFUSION\sd.webui\webui\models\Stable-diffusion\realcartoon3d_v17.safetensors
Creating model from config: C:\STABLE DIFFUSION\sd.webui\webui\configs\v1-inference.yaml
C:\STABLE DIFFUSION\sd.webui\webui\venv\lib\site-packages\huggingface_hub\file_download.py:1150: 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(
Running on local URL: http://127.0.0.1:7860/

To create a public link, set share=True in launch().
Startup time: 5.8s (prepare environment: 1.3s, import torch: 1.8s, import gradio: 0.4s, setup paths: 0.5s, initialize shared: 0.1s, other imports: 0.3s, load scripts: 0.5s, create ui: 0.3s, gradio launch: 0.5s).
Couldn't find VAE named None; using None instead
Applying attention optimization: Doggettx... done.
Model loaded in 2.8s (load weights from disk: 0.2s, create model: 0.4s, apply weights to model: 2.0s).

[ornate elk]

10 steps is to less and you need an Inpainting model for outpainting

[boreal iris]

ah alright

[fair oxide]

Hey it worked thank you do I need to upload the model or does the RX 580 work without any model like this

[ornate elk]

Looks normal.
My guess is that the civitai browser + extension needs an update to be compatible with the latest auto1111 version.

And my guess was right:
https://github.com/BlafKing/sd-civitai-browser-plus/issues/313

[smoky mesa]

You are a legend, thank you!

[ornate elk]

You can download models from Civitai.com.
1.5 and sdxl models preferable

[ornate elk]

NP, btw what's your GPU?
Your missing out on performance by not using the right launch args

[ashen phoenix]

I'm kinda confused, can I use any checkpoint models (SD 1.0, SDXL, SD 1.5, SD 2.1) with this one webui?

[smoky mesa]

4090

[ornate elk]

Then you should edit the webui-user.bat and at the line commandline_args=
You need to add: --xformers

Then save and relaunch

[ornate elk]

Yes you can

[smoky mesa]

Thank you @ornate elk 👍

[fair oxide]

I installed it but it doesn't work I got a long error message

[pulsar grail]

Would a 3070 also be better served with --xformers launch paramters?

[ornate elk]

Yes. A 3070 needs --xformers --medvram-sdxl --no-half-vae
For best performance

[fair oxide]

an exception occurred : ('Cannot copy out of meta tensor; no data!',).
While copying the parameter named "cond_stage_model.transformer.text_model.encoder.layers.0.layer_norm1.weight", whose dimensions in the model are torch.Size([768]) and whose dimensions in the checkpoint are torch.Size([768]), an exception occurred : ('Cannot copy out of meta tensor; no data!',).

[pulsar grail]

Didn't think I needed the medvram argument, ty

[ornate elk]

Its for speeding up sdxl models

[ornate elk]

Can you show me the full cmd log?

[pulsar grail]

set COMMANDLINE_ARGS= --xformers --medvram-sdxl --no-half-vae

Correct format?