🤝｜tech-support

[ornate elk]

hey, checkout #1072220168534642768 and #1080946152318443610
If you want to generate images here, you can use the discord bot for free in the bot channels: #1080261341362786384

[ivory bramble]

He is asking how to generate an image

[karmic crown]

i understand that much, but still 🙂

english is my 3rd language

but i speak english not serbian here 🙂

[ivory bramble]

Ohh it's my first, spanish is my 2nd

Nice, my best friend lives in Serbia atm

[willow zodiac]

Error in CR Multi-ControlNet Stack - 'NoneType' object has no attribute 'lower'. I'm not sure what that means

Here is the whole error message...

I have a Realistic Lineart and DWPose Estimator input to the CR M-CN Stack

[karmic crown]

lower is just a python function that put text to lowercase, but the problem is that it didn't get anything to convert, so python throw an error

that is just a consequence, we need to figure out the reason

i am to tired to think, but looks your model (or controlnet model) is missing

[willow zodiac]

There is this note in the comfy manager...

Does this mean that I can't use the controlnet preprocessors?

I guess I'm not sure what "this installation" is specifically referencing

[karmic crown]

it looks like that

but i haven used that one, so i cant be sure

i am using a1111

[willow zodiac]

ahh

[viscid flax]

hey

just finished downloading

it works now

thanks a lot for the help

[karmic crown]

if you have any further questions about prompts later, ask in https://discord.com/channels/1002292111942635562/1011743094309396631

[lavish meadow]

is it possible to make your own model?

well ofc it is but how

[karmic crown]

with kohya_ss
see if this tutorial helps
https://www.youtube.com/watch?v=RlvqEKj03WI

[silent ivy]

please help how do i solve this

[ornate elk]

When do you get that error?

[silent ivy]

When i launch it and then try to select model on stable diffusion checkpoint

[silent ivy]

When i select the model it shows error

[silent ivy]

so i m not getting is there anything i need to change

[silent ivy]

excat word wriiten there

[karmic crown]

error message is:
OSError: Can't load tokenizer for 'openai/clip-vit-large-patch14'.

it also gives a solution:
make sure 'openai/clip-vit-large-patch14' is the correct path to a directory containing all relevant files for a CLIPTokenizer tokenizer

do you have it in that place?

do you have folder openai and clip-vit-large-patch14 inside?

[silent ivy]

I do have a folder for Openai in my stable diffusion folder

[karmic crown]

and do yoi have clip-vit-large-patch14 inside

and does it have all necesary files inside (if that is a folder and not a file)

[silent ivy]

Dnt have tat one

[worldly rain]

hello, I just installed stable diffusion but I cant run it, can somebody help me out please?

[silent ivy]

What python version u using?

[worldly rain]

3.10

I installed the latest 3.12 version but then it said that 3.10 is supported

[silent ivy]

Yea

[worldly rain]

so i installed 3.10 now im getting this eror

[silent ivy]

Happened with me

[karmic crown]

you are using 3.12

please remove it if you do not really need it

[silent ivy]

Yea might uninstall and restart helps

[worldly rain]

okay i'll uninstall it

[karmic crown]

if you do, we will find to find another solution

[silent ivy]

I m stuck man

[karmic crown]

please do, it will make things easier

[silent ivy]

It's 3am dude

[quiet marsh]

hi there! anyone know what this means??

[karmic crown]

https://huggingface.co/openai/clip-vit-large-patch14/tree/main

[silent ivy]

Do i need to install all the file and which folder do i put in then

[karmic crown]

based on your error

make sure 'openai/clip-vit-large-patch14' is the correct path to a directory containing all relevant files for a CLIPTokenizer tokenizer

you need to create a folder clip-vit-large-patch14 inside
and to put all files there

but i am stabbing in the dark

i have no idea what are you doing

i am just reading your error and trying to help

[silent ivy]

Yup i can understand, let's give it a try

Look like other peoples earlier have the same issue

[karmic crown]

almost everyone here use some user interface, mostly A1111 or comfy

and we can help with those we use

but in cases like yours, we can only guess

[silent ivy]

Yup , i m downloading let's give it a try hope works

[worldly rain]

hey, I uninstalled python 3.12, I only have python 3.10 installed but it's insisting on using python 3.12

I even located python 3.10 here

[sturdy ginkgo]

delete your venv folder and relaunch

[karmic crown]

exactly that

[worldly rain]

alright I think it worked

just like that XD thank you so much

[karmic crown]

what gpu do you have? i hope nvidia

[sturdy ginkgo]

something in the venv folder saves the python directory used before

[worldly rain]

nvidia gtx1060 3gb version

[karmic crown]

ok, than it will work like that

but...

#🤝｜tech-support message

[worldly rain]

but?

[karmic crown]

find apropriate command line params for your card there

[sturdy ginkgo]

definitley need to use --lowvram in the bat file launch commands

[karmic crown]

i guess --xformers --no-half-vae --lowvram is for you

[worldly rain]

alright

is it gonna prevent me from generating hyper realistic images? or is it just gonna be slow to generate

[karmic crown]

it will make it work without errors

otherwise you probably cant generate anything

[worldly rain]

alright

I could use another pc that has gtx3060 12gb

what would be the difference?

[karmic crown]

like comparing the speed of a plane and a car 🙂

[worldly rain]

Lol alright thanks a lot

[karmic crown]

for other pc, use this guide #🤝｜tech-support message

[silent ivy]

Where I will find the Venv folder if I delete it?

[sturdy ginkgo]

will be auto made on first/relaunch

[karmic crown]

it will be created again

[silent ivy]

Lol

[silent ivy]

Noted

[rose skiff]

I'm trying to use the imag2img mode to modify an image but it's not working. The image with the transparent background is the input, the one with the white background is the output, and the prompt is "A golden retriever sitting on the moon". How do I make it follow the prompt?

I have tried both 1.5 and XL 1.0 and my hardware is a 4070 GPU, 14900KF CPU, and 64GB of ram

[karmic crown]

what is your denoise strenght?

[rose skiff]

I have tried moving the slider around a lot, right now it's 0.25

[karmic crown]

that is your problem

[rose skiff]

it's not very different at 0.67

[karmic crown]

what you get with 1?

[rose skiff]

here are my settings if that helps

[karmic crown]

what model is that?

[rose skiff]

XL 1.0

this is what it does with the same settings on normal 1.5

[karmic crown]

• for 1.5 use 512x512, 512x768 or 768x512 (that will make disfigured images)

[sturdy ginkgo]

you will have to use a mask to put the dog subject in a different background there is an extension that can do this super easy

[karmic crown]

• for xl use 1024x1024, 1024x768, 768x1024 or 768x768

[rose skiff]

How do I do that?

[rose skiff]

ok, set it to 1024 pixels tall and 768 pixels wide

[karmic crown]

one of the ways

[rose skiff]

that ui does not look like the one I am using

I have automatic1111

[sturdy ginkgo]

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

[rose skiff]

I already have a completely transparent background

[sturdy ginkgo]

that extension can generate a mask for you that you can use with the controlnet viking is suggesting

[karmic crown]

it is faster with canny + openpose 🙂

[sturdy ginkgo]

or just open gimp or whatever and select all transparent areas and invert selection fill white 😄

[rose skiff]

I think it broke it

[karmic crown]

did you changed webui-user.bat and added parameters required to your card?

[rose skiff]

When I added it it got an error about running out of memory

[karmic crown]

what is your gpu and what you have in webui-user.bat?

[rose skiff]

4070

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--xformers --autolaunch --update-check --no-half

call webui.bat

it was fine before I added --no-half

[ornate elk]

Remove --no-half

Then check the model size

[misty wigeon]

anyone had any luck with this, its been super painful to try setup

[rose skiff]

it's 6.7gb on my ssd, and it gets a different error when I remove --no-half

[ornate elk]

What's the error?

[rose skiff]

NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.

[ornate elk]

Can you show the settings you used ?

Steps, cfg, resolution etc, any loras used ?

[astral comet]

@echo off

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

call webui.bat

@ornate elk whatsup mate! it'sa me

[rose skiff]

not using any loras

just trying to draw my dog on the moon with img2img

it was drawing a copy of the original image before I added this: https://github.com/AUTOMATIC1111/stable-diffusion-webui-rembg

[ornate elk]

@rose skiff try set the resolution to 1024x1024

[rose skiff]

same error

[ornate elk]

Yes its pretty easy to install it but pretty useless for actual playing around with. Because Olive with ONNX is not very good supported.

I would recommend you installing the normal directml webui for AMD.
A guide for that is in this Channel when you checkout the Pinned Messages.

[ornate elk]

An extension active?

[rose skiff]

yes this one https://github.com/AUTOMATIC1111/stable-diffusion-webui-rembg

[ornate elk]

And when not using it it works?

[rose skiff]

when not using it it draws a copy of the original image

[ornate elk]

And you want the dog to be on the moon?

[rose skiff]

yes

[ornate elk]

Then I would suggest using inpaint

[rose skiff]

or anything other than the original image at this point

[ornate elk]

Click on inpaint. Mask the complete Dog.
Then select inpaint area "not masked"
And then prompt for the moon

Also set the resolution to 1024x1024

[rose skiff]

how do I mask the dog? I've been messing with it for a few minutes and nothing has changed

[ornate elk]

In inpaint you should be able to draw over it

[rose skiff]

I need to afk for a bit so I will have to try to make it work later

dog wants to go on a walk

[ornate elk]

Oh your dog image is with transparent background. That may not work then

Paste it into an white square image and then mask it again to test

I'm off for now.

[shy sentinel]

Hive Mind been running SD for a few weeks now want to add it on my razer laptop with egpu amd stuck with add skipcuda__test and trying to set CUDA_VISIBLE_DEVICE=1 just gets me the skip test error its trying to run on the interanal. loads but won't accept prompts cram errors out. anybody have a minute to help me with this? it would be hugely appreciated.

[karmic crown]

os?

[shy sentinel]

win11 upload screenshot now

[karmic crown]

there is a amd guide in pinned messages of this channel, but i have no idea will it work for you properly or not

[shy sentinel]

guide to switching gpu in SD ? been stuck for like an hour i am thinking its just a couple commands and it could be running.

[karmic crown]

a few weeks ago guy with amd egpu on Mac, had a completely different problems that we had on a "normal mac" with AMD

amd on windows guide -> #🤝｜tech-support message

[shy sentinel]

Copy I will check that out right now, much appreciated viking ! in MN by chance ?

[karmic crown]

i like that logo so much, but not even close to that country 🙂

[shy sentinel]

Copy ! I read what you sent in the link i am going to reset to original files and try seting visible device befor launching web ui. thanks agian !

[sonic ferry]

ComfyUI noob question, https://youtu.be/g9JXx4ik_rc?si=5aHu314EzDj4_qRm&t=889 how is he seemingly starting the queue at a certain point in the tree instead of from the beginning? 😛

[lavish meadow]

"supermerger extension" anyone got a link to this?

[zinc arrow]

weird question but is there a list of all the quotes that are written in the terminal when comfyui starts

[astral comet]

hey guys using a model, I used some prompts that gave a certain style to the images. restarted SD, used the same model, same settings, same sampling method yatti yatta, but the images now look like they're in a different style. I also did some gens with other models prior to that, anyone have an idea what's going on

[rose skiff]

thanks, it works now

now I just need to mess with the prompt

[lavish meadow]

@ornate elk Hi mate im merging models rn, do you have a standard prompt i can use to see how well it does?

[astral comet]

@ornate elk it looks like whenever i do some gens with a model, then change models, it keeps elements from the last model I used. like I was using an anime model then a realistic one and when i switched back to the anime the anime character was realistic, is that possible?

[kindred tinsel]

Hello does anyone know how to get the sd15_hand_..... model the new depth model for fixing hands

To show up on the adetailer section where it supposedly goes. Right now I am getting this None message and no option to even set it

[warm tusk]

Any idea why using the Stylez and/or StylePile addons would immediately break access via gradio.live? When either or both are enabled the connection will error out almost immediately and upon a reload gradio will say no interface is running. The local URL works fine through all of this (and does require a login/password, just like the public one).

At least Stylez used to work as of a few days.

[mental locust]

When I share a workflow can people see the list of files I see in my imageloader and videoloader nodes, or do they only see the filenames they have in their own loaders?

Think I found the answer:

Are you referring to the Input folder in the Comfyui installation folder?

Comfyui runs as a server and the input images are 'uploaded'/copied into that folder. If this is what you are seeing when you go to choose an image in the image loader, then all you need to do is go to that folder and delete the ones you no longer need.

[willow zodiac]

Can anyone help me get ControlNet preprocessors to run? I've tried on ComfyUI and now with A1111 and I have issues with both UIs.

I would prefer to run it on ComfyUI

[karmic crown]

i answered you in #📝｜prompting-help which is apropriate channel for questions like that

just a friendly request for everyone, please do not tag some people for any single question you have. i do not care if someone is tagging me or not, since i have no notifications enabled, but that might be really annoying for some other people

[karmic crown]

controlnet models goes in a different folder (ControlNet) than chekpoints (Stable-diffusion)

[karmic crown]

you need to download appropriate controlnet models and put them in a proper folder, for a1111 that is a folder ControlNet

#🤝｜tech-support message

if you have some specific issues, like which preprocessor is better for a specific use case, that is more for #📝｜prompting-help

[willow zodiac]

I tried installing this repository in the ControlNet folder then restarting A1111... https://github.com/Mikubill/sd-webui-controlnet

[karmic crown]

that is the first

than you need to download models

[willow zodiac]

When I go to the Extensions tabs, I still don't see them

[karmic crown]

do you see it here?

[willow zodiac]

I also did that from the Available tab, searched ControlNet, found sd-webui-controlnet manipulations.

no, I don't see that as an option... let me look again

[karmic crown]

that is in my list of installed extensions

[willow zodiac]

I don't have that as an option... I loaded from https://raw.githubusercontent.com/AUTOMATIC1111/stable-diffusion-webui-extensions/master/index.json

but it is not there

[karmic crown]

if you do not see it here

than installation failed

[willow zodiac]

sd-webui-controlnet manipulations is the only module tha tI have in my list

[karmic crown]

show how it looks

[ornate elk]

No its not possible. Only one model is usable at a time

[karmic crown]

the second one

click install

if installation succeeds you will see it where i see it

[willow zodiac]

I tried that and I got the follwing error

There is a file in that location when I go look, but it appears empty

[karmic crown]

remove that folder

[willow zodiac]

I deleted the folder, restarted A1111, from Available Extensions installed the sd-webui-controlnet manipulations and got this error...

and that tmp folder is back

the one I previously deleted

[karmic crown]

you can try to install it the other way

[willow zodiac]

manually from git-hub?

[karmic crown]

yes

but remove that folder first

[willow zodiac]

OK, will try that

[karmic crown]

i am going out (it's non-working day here) so i will check later, but someone else might jump in and help you even earlier

i hope it will work

[willow zodiac]

thanks

Damn...

And the tmp file is back

[karmic crown]

if you open stable diffusion webui folder from cmd and type git status what you get?

[karmic crown]

looks fine

try to restart webui and to install ANY extension from the list

[willow zodiac]

That says Version 10.0, but I have Windows 11. Is that a conflict?

[karmic crown]

civitai browser+ for example

nope, that is windows 11 obviously

they always used some "nice" version numbers since windows 95 🙈

please try to restart webui (ctrl+c first), and install any other extension

and see if that works

[willow zodiac]

I tried civitai browser+ (not uninstall/reintall) but got the same thing. I will try the un/re now 😩

[karmic crown]

wait

[willow zodiac]

yep

[karmic crown]

i would try to remove venv first

since your files from wbeui repo are uptodate

if that doenst work, try compleate reinstall

[willow zodiac]

the venv folder?

[karmic crown]

and use one of the pinned guides from this channel (realative to your GPU)

yes

inside C:\AI\stable-diffusion-webui

[willow zodiac]

deleted venv. Sorry, feel like a nube, guides pinned to the channel?

[willow zodiac]

got it. thanks

[karmic crown]

i am out

[severe coyote]

hi , can someone help me with this please?
i get this in the cmd : Already up to date.

exit code: 1

stderr:
C:\Users\User\AppData\Local\Programs\Python\Python310\python.exe: No module named pip

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

[ornate elk]

What python version do you have installed?

[severe coyote]

had 2 - 3.11.3 and 3.10.6, now uninstalled bothe, and install new one, what version i need?

[ornate elk]

Better to install python 3.10.11
And make sure to checkmark the "add python to path" when installing.
Also disable third party antivirus software

[severe coyote]

thanks, what dose it mean third party antivirus software?

[ornate elk]

Some antivirus software can bkock parts of the python installation.
Software's like McAffe, Avira, Avast, Norton etc

[severe coyote]

noe getting this : Already up to date.
Couldn't launch python

exit code: 9009

stderr:
'C:\Users\User\AppData\Local\Programs\Python\Python310' is not recognized as an internal or external command,
operable program or batch file.

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

[ornate elk]

Did you checked "add python to path" ?

[severe coyote]

yes, i also cannot delet the 3.11.3 version, i always get a fatal error with this no. : 0x80070643

[ornate elk]

Oh how did you try to uninstall it?

[severe coyote]

not in the right way i guess, i deleted the launcher before uninstall, by mistake

[ornate elk]

Yea you should uninstall it the correct way

Best would be to install python 3.11.3 again and then uninstall it again

[severe coyote]

ok. ill do that, thanks!

[ornate elk]

Correct way would be in windows over Settings>Apps or with the System Control Panel

[simple tartan]

ok, so, until we get rocm in windows, it seems i cant use controlnet on sdxl workloads.... even with 24GB of vram

RuntimeError: Could not allocate tensor with 335544320 bytes. There is not enough GPU video memory available!

[ornate elk]

On windows?

[simple tartan]

yea

[ornate elk]

Use the smaller openpose model for xl

Then it should work

Also try reducing the width to 768

[ornate elk]

Try this one:
https://huggingface.co/thibaud/controlnet-openpose-sdxl-1.0/blob/main/control-lora-openposeXL2-rank256.safetensors

[simple tartan]

ok, i have that one too

processing speed is crawling though, 18s/it

[ornate elk]

Make sure to restart before trying

[severe coyote]

done, it was uninstalled, i download the 3.10.11 version, and getting this error : Already up to date.
venv "C:\AI\stable-diffusion-webui1.6\stable-diffusion-webui\venv\Scripts\Python.exe"
No Python at '"C:\AI\stable-diffusion-webui\Python\Python310\python.exe'
Press any key to continue . . .

[ornate elk]

Delete the venv folder

Then relaunch the webui-user.bat

Oh and pls don't download the webui as zip

You can follow my install guide from the Pinned Messages in this Channel

[severe coyote]

its downloding now, seams it ok now, thanks!

[ornate elk]

If you downloaded the webui as zip you should delete it and follow my guide.
If not you will get problems in the future if you want to update SD

[simple tartan]

ok, net is running, still slow, 31.83s/it

[severe coyote]

i already had webui, i just had the python problem,

[ornate elk]

Alright, then your good 👍

[fair oxide]

Does anyone use the auto-photoshop plugin? I need help, please!

[fair oxide]

can help me ?

[ornate elk]

Depends in the issue

[fair oxide]

I'm trying to install this super easy plugin https://github.com/AbdullahAlfaraj/Auto-Photoshop-StableDiffusion-Plugin?tab=readme-ov-file#one-click-installer

but whenever I open photoshop to use it, it indicates that I didn't change the webui flag to --api, and I did it in all of them, including

[ornate elk]

Can you show me the webui-user.bat ?

Also what's your GPU?

[fair oxide]

3060

one moment

@ornate elk

I did it on all of them and it still says I didn't change it.

[ornate elk]

Dont edit any other file than the webui-user.bat

[fair oxide]

but I had only edited webui -user .bat , I edited the others to see if it worked, because editing webui -user .bat didn't work

@ornate elk

[simple tartan]

ok, images finished with those changes @ornate elk , thank you! looks like crap, but that's where the tuning starts, at least it completed

[karmic crown]

any file with .sh extension that has export in it is not for windows, those are for use on macs and linux

you need to edit webui-user.bat only (do not edit webui.bat)

[ornate elk]

Remove the changes you did in the other two files and then also add --xformers to the line with --API
Then relaunch the webui-user.bat ans show us the cmd

[fair oxide]

but that's what I said, I had only edited the webui .bat and it didn't work, it kept saying that I didn't change it to --api , so I tried the other files, so it didn't help either

[fair oxide]

ok

[karmic crown]

are you sure it is case insensitive?

i havent try to be honest

or you just wanted to emphasize that he should add --api?

[ornate elk]

No its not, I'm on Smartphone and my keyboard likes to play around with caps 🙂

[fair oxide]

@ornate elk

It didn't help, it still keeps saying the same thing

[ornate elk]

Can you show us a screenshot of the cmd after launching the webui-user.bat?

[karmic crown]

you were faster 🙂

[fair oxide]

@ornate elk oh god hahaha

It started to work because I ran the SD through webui -user .bat , as I was running it through webui.bat which in this case does not execute the commands within the webui -user .bat file

It worked

Thank you for this, I've been racking my brain over this for hours

[karmic crown]

its much easier on mac, since you just run webui.sh in both cases

[karmic crown]

on windows, you need to be sure to run webui-user.bat

[fair oxide]

Yes, because I didn't even imagine that haha

[karmic crown]

does it reuare any pluing on a1111 side, or just --api?

[ornate elk]

Perfect NP xD

[fair oxide]

requires application with --api

@karmic crown

[ornate elk]

Did you added --xformers too?
Its for better performance

[karmic crown]

i found the answer, it requires an extension:

(Don't skip) Install the Auto-Photoshop-SD Extension from Automatic1111 extension tab. The extension will allow you to use mask expansion and mask blur, which are necessary for achieving good results when outpainting and inpainting.

[fair oxide]

basically it's a folder that you put in the PS plugin, install the extension on the SD1111 with the github link and then connect the --api

[fair oxide]

yes

[fair oxide]

for controlnet

[ornate elk]

Okay good 🙂

[fair oxide]

Do you have a tutorial on how to create checkpoints and models? I'm new to SD, very layman, I barely understand anything

[ornate elk]

For getting checkpoints (models)
You should visit civitai.com
Its the largest database for SD stuff.
They have models for nearly everything.
You can also train your own model If needed. For that you should checkout some tutorials. Its not so easy.

[fair oxide]

To create my checkpoints would it be with Lora?

[karmic crown]

CS1o is typing... if file size... 🤣

[ornate elk]

A lora is not a model. Loras are trained on a specific style or character to help the model generate.
Loras dont work alone.
For training a lora you should use the Khoya_ss tool

[fair oxide]

on this civitai website, how do I do it? Do I have to download the models or something like that and then where would I have to go, if that's the case?

[ornate elk]

For example download the Dreamshaper v8 model from civitai.
Then put it into the models/stable-diffusion folder

[karmic crown]

i think that you can use Khoya_ss to train a model too (but i might be wrong)

[ornate elk]

Yes you can

But its not easy xD

[fair oxide]

so easy ?

[ornate elk]

You have to reload the checkpoint dropdown in auto1111 after adding the models

[karmic crown]

i dont think it is too complicated, but it would probably take forever on my hw 😦

[fair oxide]

how do you do this?

[ornate elk]

True

[ornate elk]

When you launch the webui-user.bat it should also open a browser tab for the webui

There in the top left corner

Is the checkpoint dropdown

And a refresh button

[fair oxide]

I'll download one here to test

[fair oxide]

You need to log in based on what I'm seeing

[ornate elk]

In civitai?

You dont need an account, only for nsfw stuff

[karmic crown]

or for creating images

[fair oxide]

It was necessary to log in, it was not possible to download the model

but I did that super easy procedure, putting the model in the models/SD folder and opening the SD here and it is working perfectly. What happiness hehe

[ornate elk]

Some models require a login that's right. But the majority not

[karmic crown]

there is even esier way, install civitai browser+ extnsion

[ornate elk]

Perfect 🙂 if you try out different models check what base version they are.
Some are 1.5 and some sdxl for example.
1.5 models are trained on 512x512 resolution and sdxl on 1024x1024.
Keep that in mind when you use them in Photoshop

[karmic crown]

it will download model and put it in right folder

[fair oxide]

oh yes, true there is still that detail

[fair oxide]

Thanks for the tip, but I don't think it will be necessary

@ornate elk @karmic crown If you use auto-photoshop one day, I did the test here and the webui-user.bat didn't work when using the SD1111. I went to generate the images and everything was gray, it only serves as a plugin for photoshop. I ran webui.bat and it worked normally

[ornate elk]

So does it work now or not?

Only ever run the webui-user.bat

[fair oxide]

I meant that webui -user .bat is not working to run SD1111, everything is bugged

webui -user .bat = PHOTOSHOP PLUGIN

webui.bat = SD

[fair oxide]

Do you have any tutorials on how to create models?

[karmic crown]

it should work from webui with webui-user.bat

actually IT MUST work

[fair oxide]

I separated the two executables, one serves only as a plugin and the other runs the SD

[woven smelt]

Hello i want to use the refiner, before I always loaded the same model as I was using but it says "its base model use refiner extension disabled" Should I just use sd_xl_refiner instead?

[silent ivy]

That was a clutch bro , it worked after i downloaded those file

[karmic crown]

i wish i could fix my problems with download 🤣

[woven smelt]

Found that the sdxl refiner crashes if used with controlnet

[silent ivy]

Deep

[karmic crown]

link + user + password on some publicly accessible webui with a100 would also solve my stable diffusion problems 🤣

[silent ivy]

Is their any history in stable diffusion I mistakenly generate new lol

Nvm I made the image better

What are the best model ? I can use @karmic crown

[karmic crown]

open https://civitai.com/ and find the one with gallery that you like the most

i pesonally like https://civitai.com/models/241445 and https://civitai.com/models/119012

[silent ivy]

Noted ,need a guide I think on those tools let me download

[karmic crown]

since you are using code, you need to download a model, but also to change your code

you should use one of those instead of base sdxl model (sd_xl_base_1.0.safetensors)

[fair oxide]

I didn't even know you could create nsfw with stable diffusion

[simple tartan]

seriously? that's like 85% of the userbase, lol

[fair oxide]

hahahaha

[karmic crown]

once i had a problem with a "normal" model with a specific seed, i added nfsw, nude, naked, topless... in negative prompt, but i always got a topless girl, even after I added weights like (nsfw:1.5)

[fair oxide]

These guys are really horny lol

[silent ivy]

Let me try then , i think u need to write promet only

[silent ivy]

Noted

[fair oxide]

?

[karmic crown]

it will not work for you as long as you are using default sdxl model

sd_xl_base_1.0.safetensors was trained on swf images only

[silent ivy]

I have another model realistic model

It worked

Just created

Hope i dnt get ban lol

[karmic crown]

why would you get banned? you didnt post image here, and you didnt insult anyone

[silent ivy]

No like on stable diffusion app

[silent ivy]

85?? Bruh

[fair oxide]

How do you create models? Do you have a tutorial to share with your friend here? +)

[karmic crown]

start with something easier, like lora

[silent ivy]

No i downloaded 2 model one was realistic model another was default one

[silent ivy]

Their is a section called lora , but nothing there

[karmic crown]

this is lora for example https://civitai.com/models/11722/iu

[fair oxide]

I'm going to download Lora through Pinokio

[silent ivy]

Let check yt tutorial on that

[karmic crown]

loras chan afect style, face or level or details

[silent ivy]

ur UI looks different

[karmic crown]

oops, i misread 🙂

[silent ivy]

like this section

[simple tartan]

you have to download and install them first 😉

[silent ivy]

mine like this

[karmic crown]

yes i have custom theme for a1111, sorry, i thouth you were talking about IU (Korean singer and acctress) 🙂

[silent ivy]

IU my fav

[fair oxide]

How do you configure the interface?

[silent ivy]

i saw her kdrama

[karmic crown]

go to extensions tab, load available and search for lobe theme

install and relaunch

[fair oxide]

oooooh

amazing

[karmic crown]

i have blue on my sons pc, and green on my mac, so i know which one is which without looking at the address bar 🙂

[fair oxide]

son ?

what your year ? haha

age*

[karmic crown]

my son is 21, so guess 🙂

[fair oxide]

hahah

hmm

45 ?

[karmic crown]

recently i found my business card from last century 🙂

[fair oxide]

Assuming you get it early

[silent ivy]

nothing here dude

[karmic crown]

click available tab

[silent ivy]

i m 25

[karmic crown]

click load from button, and search for lobe theme

found it?

[silent ivy]

now?

[fair oxide]

there are many

which one

[karmic crown]

no

search for "lobe theme"

[silent ivy]

okay

wait

[karmic crown]

for loras, go to civitai

[silent ivy]

i didnt downloaded yet

[fair oxide]

I put lobe theme but many options appear

[karmic crown]

https://github.com/lobehub/sd-webui-lobe-theme

[silent ivy]

okay its downloaded

[fair oxide]

Stranger doesn't appear to me

[karmic crown]

when you install extension, you need to restart ui

[silent ivy]

i m cloning

[karmic crown]

but in some (rare) cases, it is best to restart the whole ui, so ctrl+c
and launch it again

[silent ivy]

u can see lora there

shall i untick those build in

[karmic crown]

lora queue helper is something we didnt talk about

[silent ivy]

oh lol

downloaded the wrong one i guess

[karmic crown]

we talked about loras (in general) and my custom user interface

if you want my user interface you need to install lobe theme extension

if you want loras, go to civitia.com and find some you like

put them in modes/lora folder inside your installation

[silent ivy]

W

[fair oxide]

I downloaded it here, it was good, but it's a bit confusing... maybe I'll have to get used to it

[silent ivy]

THIS one right

[karmic crown]

just remember to scroll up and down to navigate trough main navigation, than it will be easy

[karmic crown]

for lora usage, always check the info on page where you download it, and examples on the same page. if you still have problems with lora usage, ask for help in #📝｜prompting-help

[kindred tinsel]

Hello I am back here once again to ask for help aobut a weird ass error I am getting

Basically I downloaded everything and pretty much have all the models isntalled, but I tried running it through an image and it threw this error immediately

[ornate elk]

what are you even trying to do xD

i see controlnet

[kindred tinsel]

I am trying to use the new hand depth model, its only to fix up hands, I was trying to see how it masked stuff and try to use it later on when in inpainting to fix the hands using that controlnet and a prompt, but the test failed with that

[vocal burrow]

screenshot your whole webpage

[kindred tinsel]

there you go, it shows the error on the controlnet, I can send the error message to, although its the same as the txt I sent

[vocal burrow]

reload your webui, check the " allow preprocessor" in controlnet and before generating click the "explosion logo" so we can see what the preprocessor gives

also is the model you re using sd15 based ?

[kindred tinsel]

yeah

also

I tried doing that, im running some images without the controlnet for hands so im not sure if I can stop it rn

[vocal burrow]

and only activate one controlnet module for now

[kindred tinsel]

Rn yeah I am running only one, yesterday I ran like three togther so Im sure its not neccesarily other preprocessors that give me these issues

[vocal burrow]

when there some error, you always isolate it. It s easier and faster to work that way

[kindred tinsel]

Fine fine give me aprox 2 minutes to finish adetailer pass and then Ill rerun evertyhing, webui and all to test out only that one error

@vocal burrow Base settings, want me to increase preproccesor resolution maybe?

[vocal burrow]

"allow preview" and click the explosion logo

[kindred tinsel]

Okay midas passed, but hand refiner gives error

[cyan fable]

Hi everyone! I'm trying to get fooocus to work, is this the right channel?

[kindred tinsel]

Yep it is !!

Also it says the backend might not be allowed, like propbably amd is not available as a backend for this thing, do you think I could change the backend

[ornate elk]

tried the hand refiner, got the error too

[vocal burrow]

why are you using an inpainting controlnet model ?

[kindred tinsel]

inpainting where?

Im using depth, like normally you would use, to at least get the mask

[vocal burrow]

inpaint depth hand

[kindred tinsel]

literally the model that was given out in the hugging face thingy

I might be dumb and using it in the wrong place but idk im trying now in the impaint tab

[vocal burrow]

can you try with the control_v11p_sd15_depth ?

[kindred tinsel]

Should I try with a mask over the hands or just let it do its thing?

[vocal burrow]

(it s all relatively news so I wouldn t be surprised if the main version is broken, it might be necessary to switch to the dev branch)

[vocal burrow]

try like you just did but using control_v11p_sd15_depth model and depth_hand_refiner preprocessor

[kindred tinsel]

im on it ^^

[cyan fable]

okay good! so I followed a video tutorial to download fooocus and I've managed to get the program to generate images, but I'm having some issues. The tutorial showed that my checkpoints folder is supposed to have sd_xl_base and sd_xl_refiner files that are downloaded when you first run the program, but they've not appeared, even after re-installing. Also, it's taking an incredibly long time to generate photos, 5-10 minutes per photos at 7 seconds per iteration when the github page states that a low-end laptop can do 1.5 sec per iteration. My computer specs are Nvidia GeForce GTX 1660 SUPER, 16 GB 3200MHz RAM, AMD Ryzen 5 5600G and installed on a 1T SSD drive.

[vocal burrow]

I cant test stuff right now I m at work and will probably be very busy tonight, so I ll just give you some resources to take a look at https://github.com/Mikubill/sd-webui-controlnet/discussions/2400

[kindred tinsel]

It returned the smae image no changes, kinda weird but it still gives me the error,

Its weird but thanks for trying ^^

[vocal burrow]

I might have time to try it out on dev branch later

[ornate elk]

hey, what model do you currently use?

[dense tapir]

@ornate elk

[cyan fable]

let me check

[dense tapir]

It seems like I can only manually install extensions 🤔 but not from the UI. I also tried fixing the dubious ownership with ChatGPT but seems like it still persists

[cyan fable]

it's the latest version, filename is Fooocus_win64_2-1-831.7z

[ornate elk]

you installed Auto1111 webui the wrong way (by downloading it as zip). When did you installed it?

[ornate elk]

no i mean which model do you use to generate images?

[cyan fable]

oh! uh Juggernaut

[vocal burrow]

I bet your g: partition is fat32 and not ntfs

[dense tapir]

Like an hour or two ago, I'm trying to move it to my new portable ssd

[cyan fable]

i run the realistic one

[dense tapir]

exFAT

[ornate elk]

Juggernaut or juggernaut xl ? because XL models are larger and slower

[vocal burrow]

yeah so just like fat32 it doesnt have any support for permissions / ownership

[cyan fable]

I have this file in my checkpoints folder 🤔 juggernautXL_version6Rundiffusion.safetensors

[dense tapir]

Crap, okay. Thanks

[cyan fable]

so the XL

[dense tapir]

I had it as exfat because that was the only way to use it on a phone too :/

[cyan fable]

I'm not chosing anything when I try to generate an image though, just typing in the prompt and hit generate

[ornate elk]

yea thats why its so slow. with your gpu you should stick to 1.5 models (the ones that are 2-4gb) not 6gb

[cyan fable]

oh, okay! that's what was downloaded by default so I assumed it would be fine

i can find smaller models on civit ai?

[ornate elk]

yes, there is also Juggernaut Aftermath

its 2gb

[vocal burrow]

So yup, use a proper filesystem with permissions / ownership support and it should be fine. Otherwise you can try to force git to ignore those by running something like git config core.fileMode false (run it in your stable diffusion folder of course)

[dense tapir]

Idk maybe I was just stupid as CS1o said and installed SD incorrectly, could you please link me a tutorial on how to correctly install it?
I'm not sure if this is relevant, but I already have it on my main SSD and I'm simply trying to get it on my other on as well.

[cyan fable]

okay, so I want to download this to the checkpoints folder and get rid of juggernautxl?
https://civitai.com/models/46422/juggernaut

[formal trout]

Hi, I'm getting black mask when using these inpaint SD XL ControlNet models, any clue? https://civitai.com/models/136070?modelVersionId=271562 I'm on Automatic1111, Ubuntu, RX 6700, using --medvram-sdx --opt-sub-quad-attention --opt-sdp-attention --no-half-vae

[vocal burrow]

I mean he also is right. I see that master in your stable-diffusion-webui-master folder... meaning you downloaded the whole thing as a zip probably instead of using git or the release archive.
So yes you should probably reinstall it properly. But it will also probably not fix that specific issue.

[dense tapir]

Like this?

[vocal burrow]

try using --no-half on top of that or instead of --no-half-vae

it s gonna increase your vram usage by a lot but it might fix it

[vocal burrow]

yes

[dense tapir]

I don't think it worked

[vocal burrow]

and check pins for install guides

[ornate elk]

yes that one will be faster

[dense tapir]

Thank you!!! 🫶

[formal trout]

great thanks, I'm trying it

fp16 model may be better?

or fp32 should work?

[vocal burrow]

it shouldnt matter for that.

[dense tapir]

So would you say that I should reformat by SSD then?

Before installing SD on it?

[vocal burrow]

maybe try a proper reinstall before that

[dense tapir]

Okay!

[cyan fable]

so i have the regular run.bat, and then a run_realistic.bat. will they both use the better juggernaut model and be quicker? realistic downloaded another checkpoint called realisticStockPhoto_v10.safetensors when i first ran it

[ornate elk]

these are presets (so it loads a model thats realistic) but you can change the model to your liking in the webui

[cyan fable]

okay! bless you for the help, im not too great with computers

I got an error - NotImplementedError: You have selected base model other than SDXL. This is not supported yet.

[ornate elk]

hmm, i thought foocus supports 1.5 models

can you set the styles to None?

[cyan fable]

same error :(

[ornate elk]

can you try close it again and start the run.bat not the realistic one?

[cyan fable]

I have been running the run.bat

[ornate elk]

oh okay

[cyan fable]

I'm trying again with realistic, using that model

[vocal moth]

Does someone know if its possible to use control net combined with multi diffusion and if yes is there something i need to know? For example ipadapter doesnt work at all when i tried it sadly

[ornate elk]

so after doing a bit of research, Fooocus mainly supports SDXL and is testing 1.5 support.

[ornate elk]

i know it works with Tiled VAE, but didnt tried with Multidiffusion, do you get an error?

[dense tapir]

`raise RuntimeError("\n".join(error_bits))
RuntimeError: Couldn't checkout {name}'s hash: {commithash}.
Command: "git" -C "G:\AI\stable-diffusion-webui\repositories\stable-diffusion-stability-ai" checkout cf1d67a6fd5ea1aa600c4df58e5b47da45f6bdbf
Error code: 128
stderr: fatal: detected dubious ownership in repository at 'G:/AI/stable-diffusion-webui/repositories/stable-diffusion-stability-ai'
'G:/AI/stable-diffusion-webui/repositories/stable-diffusion-stability-ai' is on a file system that does not record ownership
To add an exception for this directory, call:

    git config --global --add safe.directory G:/AI/stable-diffusion-webui/repositories/stable-diffusion-stability-ai

Press any key to continue . . .`

I'm gonna run that command and get the same kind of errors

[ornate elk]

@cyan fable for your GPU it would be the best to install the Automatic1111 webui. Its always slow to run 6gb models on a 6gb vram gpu but on Auto1111 you could use the smaller models and improved performance aswell as more features in general.
I made an hopefully easy to understand Install Guide here at the Pinned Message of this Channel.
Try that, and for any questions feel free to ask here 🙂

[cyan fable]

the realistic model ran better, 5.8s/it, but still super slow. so i should forgo fooocus entirely? dang, i spent all day trying to get this thing to work lol

[ornate elk]

Yea its made for sdxl and higher end GPUs it seems.
Your better with auto1111
Its the most used webui here so plenty of helpful people too

[dense tapir]

my SD folder is not a git directory 🤔

[ornate elk]

Because you installed it the wrong way (by downloading it as zip)

[dense tapir]

I followed your tutorial this time

The pinned message

[ornate elk]

Did you moved the folder afterwards

?

[dense tapir]

No

[ornate elk]

Can you show me the error?

[dense tapir]

Here

[ornate elk]

Is that on your external drive?

Because it says problem with ownership

[dense tapir]

It's a portable SSD, not the built in one

[ornate elk]

Best thing is to reformat it and then try again. Make sure its ntfs

[dense tapir]

Thanks! Willy try

[ornate elk]

But save your personal stuff before formatting

[astral comet]

hey guys, any way to see seeds from previously generated images, same way SD has a folder with past gens?

[dense tapir]

Put in into the png info inside the UI

Like drag the image into it

[astral comet]

in img2img?à

[dense tapir]

Here!

[astral comet]

yoooo thanks dude

[dense tapir]

😉 👍

@ornate elk @vocal burrow Thank you guys! Seems to be working. If not, then you know the drill

[ornate elk]

Perfect np 🙂

[vocal moth]

Yeah when i turn on tiled diffusion and have ip adapter selected in control net i get the following error:

File "F:\AI\AUTOMATIC1111\stable-diffusion-webui\extensions\multidiffusion-upscaler-for-automatic1111\tile_methods\abstractdiffusion.py", line 499, in prepare_controlnet_tensors if len(control_tensor.shape) == 3: AttributeError: 'dict' object has no attribute 'shape'

[sleek breach]

Hey guys, I just wanted to say, for how long should the "Installing requirements for CodeFormer" part of the bat file occur? Cause it's been there for me for a good 5 minutes

[ornate elk]

Hey, can you show a screenshot?

[sleek breach]

Actually

Now I'm getting this error

I never used discord so it's a bit hard to get used to

[vocal burrow]

before closing it can you show the full log of the installation (copy paste the whole thing in a .txt and drop the file in here)

[dense tapir]

Hey, I'm getting infinite loading times now.

[vocal burrow]

check for errors in console log

[ornate elk]

What does the cmd shows?

[dense tapir]

No errors 🤷‍♂️

[ornate elk]

And where does it load? When generating? Or selecting a model?

[pure forge]

How fast should an m2 macbook generate on sdxl turbo in automatic? I'm getting 1 minute 25 seconds when I prompted with simply "spaceman"

[dense tapir]

Wait what the

Ok

So

So I got this error:

And then my keyboard and mouse died 😹

[ornate elk]

Did you download multiple files at the same time?

[dense tapir]

I downloaded some extensions yeah

[vocal burrow]

weird I don't see anything wrong.... Is this a fresh install ?

[vocal burrow]

depends of the settings you've used

[silent ivy]

@karmic crown how do you generate prompt? I wanna create an Ai model , but prompts are not that great

[pure forge]

https://media.discordapp.net/attachments/211187501153386496/1193990337010606151/image.png?ex=65aeb927&is=659c4427&hm=47a2fb2ddaa3154262bda685a0d5e7cef3ebae81be51883d2f9cb46815606419&

[sleek breach]

Kinda, I had stable diffusion installed on another PC, I copied and pasted the files from there in here, in order to get the Loras and models here as well, maybe that could influence it?

[pure forge]

Here are the settings

[dense tapir]

Ok I need to restart my PC some of the usb ports died lmao

[vocal burrow]

yup don't copy paste from another computer. There's many things that could go wrong doing that

Do a clean install instead (and then copy paste models, loras, etc)

[proper pike]

I am using a pixel art lora from civit and when I do the native 1024x1024 model it gives weird glitches like this, but if I use hires fix 2x the pixels are perfect but the proportions are all wrong. Anyone know how to fix this?

[vocal burrow]

it will probably fix everything

[proper pike]

using hires fix and it has weird proportions on all the outputs

[dense tapir]

Upscale via scripts

[crimson sail]

hello, i am new to this whole ai thing and i installed the automatic 1111 webui, but no matter what propts i use, i get a black square. Can someone help me?

[proper pike]

How do you mean via scripts?

[vocal burrow]

1/ that seems a little slow, but the last mac I owned was an apple2 so I don't got any proper point of comparaison
2/ that's not how sdxl_turbo is supposed to be used, you're meant to set CFG scale at 1 and sampler steps at 1 (or use this extension, https://github.com/seshelle/CFG_Rescale_webui with CFG scale at 7 + CFG Rescale at 0.7 + sampling steps at 1)

[dense tapir]

So uhh what does this mean? I simply opened webui.bat

[pure forge]

Danke

[vocal burrow]

it means it's starting up ?

[dense tapir]

Oh so I just need to wait

Damb

[vocal burrow]

what does the console log say ? what settings are you using ? what's your hardware ?

[dense tapir]

I'm fun

[crimson sail]

do you mean gpu?

[vocal burrow]

it should start printing something else pretty much immediately, make sure the cmd windows is not paused / has the focus

[vocal burrow]

among other things yes

[dense tapir]

The startup time took 180 seconds for some reason

[vocal burrow]

better using those scripts https://github.com/Coyote-A/ultimate-upscale-for-automatic1111

[dense tapir]

Yes

[dense tapir]

This better @proper pike

[crimson sail]

gpu: geforce GTX 1650 and cpu: Core I5-9400F

[vocal burrow]

ok then what about the rest. what does the console log says ? what settings are you using ?

[crimson sail]

with console log, do you mean my cmd?

[vocal burrow]

yes

[vocal burrow]

don't use --disable-nan-check

[crimson sail]

but then it doesn`t work

[vocal burrow]

and it doesn't work with it either

[proper pike]

But do you know why the native 1024x1024 has weird artifacts?

[vocal burrow]

like the title says it's only gonna disable some check.... So instead of having a meaningful error message the program is just gonna ignore any errors and poop its pants without telling you anything useful.

[crimson sail]

ok i`ll disable it

if i try it now it just says:

[vocal burrow]

use --no-half

[crimson sail]

ok 1 sec

[vocal burrow]

it's mandatory for 16XX gpus

[dense tapir]

Yeah it's still horrendous, I'm simply trying to change models and the processing is at 370 seconds

[crimson sail]

btw is that a problem? "Style database not found: X:\AI\stable-diffusion-webui\styles.csv"

[vocal burrow]

no, the file should create itself the moment you save a style

[crimson sail]

ok

[dense tapir]

If I understand correctly, then I think that if the model is not an XL one, then that kind of resolution will not work because they aren't optimized for it

[fluid mountain]

what happend to deepfloyd? :( will it come back??

[crimson sail]

ok it works thx : )

[dense tapir]

Ye ok I waited 666 seconds Idk

[vocal burrow]

what's the problem exactly (kinda got lost with all the noise) ?

[dense tapir]

Lol you're good. Just infinite loading times

And then best i can do is just close SD

[vocal burrow]

that I got but when ?

what are you doing to trigger them ?

what's in the log ?

[dense tapir]

If I remember it's still nothing

I'll boot it up quickly

[vocal burrow]

looks good (apart from the lack of command line arguments so you're probably leaving perfs on the table there)
startup time 20.1s

[proper pike]

This is the lora https://civitai.com/models/120096/pixel-art-xl

[dense tapir]

What's the checkpoint?

[pure forge]

hmmm faster but not dramatically so, most of the time is spend without any progress being shown then the progress bar goes by super fast

[proper pike]

sdxl base model

[dense tapir]

Spam generate then I guess

Maybe like a batch of 15

[proper pike]

I did batch 50 lol

[dense tapir]

wtf

All of them looked like that?

[vocal burrow]

waht does the log says ?

[dense tapir]

Damn...

[proper pike]

99% are elongated torsos.

[dense tapir]

Bad lora?

[vocal burrow]

probably bad settings, prompts, etc. Try asking for more help in #📝｜prompting-help

[proper pike]

This is with hires fix, right? So thats when it has weird torso or features. However if I keep it at 1024x1024, it keeps proportions BUT the image gets weird artifacts. As if its not a pixel image

[dense tapir]

Prompt

[proper pike]

1024x1024 has weird artifacts

[vocal burrow]

again, looks like incorrect settings, try to play with the cfg scale, weight of loras, etc #📝｜prompting-help

[dense tapir]

Best results without using, pixel art in the prompt

Whatever that is

Hope you're using it

[pure forge]

how do I view that?

[vocal burrow]

it's printed in the console / cmd

[pure forge]

Total progress: 100%|███████████████████████████████████████████████████████████████████████| 1/1 [00:07<00:00, 7.35s/it]

this time 38 seconds so a little faster

[vocal burrow]

what's the FULL log

from the beginning

[dense tapir]

Yeah seems like SD completely shitted for me...huh

Not sure why Imma just leave it. Thanks for the help though! @vocal burrow

[vocal burrow]

I'm still waiting on your reproduction steps tho. What's causing the problem ?

[dense tapir]

I launch it

Lol

[vocal burrow]

and then what ?

cause it launches under 20 secs

[dense tapir]

And then if I do anything that requires the little loading animation it stucks

[vocal burrow]

try cleaning your browser cache

[dense tapir]

Okay

[vocal burrow]

if it does nothing delete the venv folder

still nothing, clean reinstall just to make sure.

[vocal burrow]

You're using python 3.9.6, you're lucky it even runs with that. SD is meant to be used with 3.10.6 up to 3.10.13.

Try reinstalling following this guide https://github.com/viking1304/a1111-setup/discussions/2

[pure forge]

yeah I missed that step smh

alright I upated that and pip and then reran

@vocal burrow damn now it took over 2 minutes

[vocal burrow]

did you reinstall following the tutorial ?

[dense tapir]

Yeah still bad...I'll just reinstall another time. It seemed like when I started getting my extensions, that's when SD just shitted

[vocal burrow]

delete them (they should be in extensions folder) and clean venv once again to confirm

[dense tapir]

Already did :/

[vocal burrow]

did you clean the venv after deleting them ?

[dense tapir]

Yes

[vocal burrow]

:/ ok maybe the reinstall will fix it. I have to go for now

[dense tapir]

Bye and thanks!

[pure forge]

I just updated python using the command in the tutorial since it's the only thing I didn't do before and then updated pip, do I need to reinstall everything after updating python?

[vocal burrow]

I'd say a clean reinstall would ensure that you're not piling up on errors from the past.

Plus if you changed your python without deleting the venv folder it's still using the outdated venv

[pure forge]

[pure forge]

Idek what venv is and Google's no help

[vocal burrow]

the venv folder in stable diffusion's folder

it basically holds a copy of everything python related necessary to run SD

that way instead of messing with your system's python files, it messes with things within its own little sandbox

venv , virtual envrionment

[sleek breach]

Well, I'm having this issue now

C:\Users\Alberto\Documents\Stable Diffusion\stable-diffusion-webui>git pull
Already up to date.
venv "C:\Users\Alberto\Documents\Stable Diffusion\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: v1.7.0
Commit hash: cf2772fab0af5573da775e7437e6acdca424f26e
Installing torch and torchvision
C:\Users\Alberto\Documents\Stable Diffusion\stable-diffusion-webui\venv\Scripts\python.exe: No module named pip
Traceback (most recent call last):
File "C:\Users\Alberto\Documents\Stable Diffusion\stable-diffusion-webui\launch.py", line 48, in <module>
main()
File "C:\Users\Alberto\Documents\Stable Diffusion\stable-diffusion-webui\launch.py", line 39, in main
prepare_environment()
File "C:\Users\Alberto\Documents\Stable Diffusion\stable-diffusion-webui\modules\launch_utils.py", line 378, in prepare_environment
run(f'"{python}" -m {torch_command}', "Installing torch and torchvision", "Couldn't install torch", live=True)
File "C:\Users\Alberto\Documents\Stable Diffusion\stable-diffusion-webui\modules\launch_utils.py", line 116, in run
raise RuntimeError("\n".join(error_bits))
RuntimeError: Couldn't install torch.
Command: "C:\Users\Alberto\Documents\Stable Diffusion\stable-diffusion-webui\venv\Scripts\python.exe" -m pip install torch==2.0.1 torchvision==0.15.2 --extra-index-url https://download.pytorch.org/whl/cu118
Error code: 1
Presione una tecla para continuar . . .

Couldn't install torch

[vocal burrow]

What do where python, python --version and py --list-paths say ? (run those in cmd)

[sleek breach]

For where python: C:\Users\Alberto\AppData\Local\Programs\Python\Python310\python.exe
For python -- version: Python 3.10.6
For py --list-paths: Installed Pythons found by py Launcher for Windows
-3.10-64 C:\Users\Alberto\AppData\Local\Programs\Python\Python310\python.exe *

[vocal burrow]

what's in the C:\Users\Alberto\AppData\Local\Programs\Python\Python310\Scripts folder ?

[sleek breach]

pip.exe
pip3.10.exe
pip3.exe

That's what I got

[vocal burrow]

do you have those in your stable-diffusion-webui\venv\Scripts folder ?

[astral comet]

hi guys, followed this tutorial, and when i select the .cpk file as a model it just glitched out and automatically switches to my last used model, any idea what could be causing this?

[sleek breach]

Nope, Only got python

And pythonw

[vocal burrow]

ok how much space do you have on your c: drive and whatever drive you have installed SD on ?

[sleek breach]

107 GB I got

[vocal burrow]

ok delete the venv folder and try to launch webui-user again

[vocal burrow]

what software are you using ? what .ckpt ? what do you means "it gliched out" ? what does the log says ?

[astral comet]

it's a custom .ckpt i made following a tutorial to build your own model

let me get back to u with the log

[pure forge]

I deleted venv and tried to git clone and it didn't reappear. How do I reinstall?

[vocal burrow]

git clone the venv folder ?!?

what

the venv folder will reconstruct itself upon running webui-user

[pure forge]

no the whole automatic

I was just using that to see if it actually updated it

[vocal burrow]

ok ... so you deleted your whole stable-diffusion-webui folder ? and git cloned it again ?

[ornate elk]

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

[vocal burrow]

hey ! spoiler ! that was my next step.

they vanished anyways :p so maybe rebuilding the venv was enough

[pure forge]

I only deleted venv and got cloned automatic and venv didn't show up again so I don't think git cloning did anything

[vocal burrow]

you don't update with git clone but with git pull

so I have no idea what your git clone did because it depends what you folder you ran it from

[pure forge]

I did git pull first and it said it was up to date

[vocal burrow]

Anyways like I said you don't "reinstall venv"

you simply run webui-user again and it will rebuild itself

[pure forge]

I didn't try to reinstall venv

Okay so delete venv and then run it?

Before you said not to just run it and that it had to be reinstalled

[vocal burrow]

< wait what.... so you didn't delete stable-diffusion-webui.... you git cloned it AGAIN somewhere.... so presumably you might have it twice.... And you're asking me how to reinstall "it" again ?!?

I'm lost

[vocal burrow]

I said that it would be better if you reinstalled it yes. Meaning if you reinstalled it completly, uninstalled all that you did, delete your stable-diffusion-webui, your python install, etc

[pure forge]

Wait I need to delete Python from my system?

[vocal burrow]

And then you follow the tutorial I've sent you. This way we're sure you're following an approved and tested procedure.

[pure forge]

Is there a step by step process for reinstalling somewhere

[ornate elk]

yes, in the Pinned Messages of this channel is a all in one install script, that also makes the needed settings for best performance on mac

[vocal burrow]

1. delete the whole stable-diffusion-webui folder
2. make sure that you have a proper and compatible python, if not uninstall it and install a proper one (3.10.6 to 3.10.13)
3. install stable-diffusion-webui following the guide provided.

[pure forge]

So just run that script and it'll ensure everything is fresh and fixed

[vocal burrow]

I've just stated what you have to do

[pure forge]

So you think the script is a bad idea

[vocal burrow]

I think it's step 3

and you shouldn't skip step 1 and 2

[ornate elk]

yea the 3 steps areyetis told will do it, the script will do the rest

[formal trout]

I did but now it runs out of memory, I tried with Low Vram in ControlNet

OutOfMemoryError: HIP out of memory. Tried to allocate 1.04 GiB (GPU 0; 11.98 GiB total capacity; 11.11 GiB already allocated; 568.00 MiB free; 11.30 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_HIP_ALLOC_CONF

Also I'm getting

ControlNet - WARNING - A1111 inpaint and ControlNet inpaint duplicated. ControlNet support enabled.

I'm inpainting with a SD XL inpainting model and I did some modification in configuration files, don't know if that is related

[vocal burrow]

AMD and inpainting ain't working pretty great.... I'm not an AMD guy so I don't know the exact commands that can make it work for sure for every single cards out there.
Usually the rule is if inpainting f*** up with AMD hardware => go --no-half
So maybe you can try --lowvram --no-half --opt-sub-quad-attention --opt-split-attention-v1

it will be even slower than before tho :/

[formal trout]

well I just tried with --lowvram ( --lowvram --opt-sub-quad-attention --opt-sdp-attention --no-half-vae --no-half ), it started but with a black masked, and awfull slow like 9s/it

[vocal burrow]

@ornate elk do you remember the magic amd values ?

[formal trout]

inpainting with regular SDXL models and the inpainting model worked for me with --medvram-sdx --opt-sub-quad-attention --opt-sdp-attention --no-half, but with this ControlNet inpainting model it seems doesn't like it

I wanted to fly too high xD

[ornate elk]

Its --medvram --opt-sub-quad-attention --opt-split-attention --upcast-sampling --no-half-vae

[ornate elk]

Use the smallest xl controlnet files and enable lowvram mode in controlnet

[formal trout]

let me try, I'm trying these one https://civitai.com/models/136070?modelVersionId=271562

[ornate elk]

Yea these are to large

[formal trout]

is there other inpainting CN model for SD XL?

[ornate elk]

I dont think so

[formal trout]

yes me neither, it was the first one I saw xD

[ornate elk]

small sdxl controlnet files can be found here, but no inpainting
https://huggingface.co/lllyasviel/sd_control_collection/tree/main

[formal trout]

it generate but black mask sadly

[ornate elk]

--no-half is still needed for inpainting

[formal trout]

ty anyway, I'm happy with these inpainting SD XL model I found https://civitai.com/models/245423/juggerxlinpaint

But about the size, I run different CN models and a few at the time (only had to use Low VRAM when using like 4 or 5), and they worked

Some OpenPose or Depth model that were like 6 GB

[ornate elk]

Havent seen any controlnet model larger than 5gb

[formal trout]

yes 5 GB

[ornate elk]

If you can always use the smaller ones

But yea big ones can work too, make sure to enabled tiled vae too

Maybe not using hires upscale by 2

[formal trout]

that one worked without changing anything, but well these ones I don't know, I guess there will be more in the future

[ornate elk]

Yea I think so too

[formal trout]

some guy say this but it didn't worked for me

got

but well as I said those new juggernaut inpainting models seems to work pretty well

[astral comet]

hi guys, so I used DreamBooth to create a model based on some pics, went through the whole tutorial which ended in downloading the created .ckpt file. I put in in the models folder, and when I try to load it it sends me back to the last model used and it says

changing setting sd_model_checkpoint to SP.ckpt [125ff6233c]: AttributeError
Traceback (most recent call last):
File "C:\stable-diffusion-webui\modules\options.py", line 146, in set
option.onchange()
File "C:\stable-diffusion-webui\modules\call_queue.py", line 13, in f
res = func(*args, **kwargs)
File "C:\stable-diffusion-webui\modules\initialize_util.py", line 174, in <lambda>
shared.opts.onchange("sd_model_checkpoint", wrap_queued_call(lambda: sd_models.reload_model_weights()), call=False)
File "C:\stable-diffusion-webui\modules\sd_models.py", line 773, in reload_model_weights
state_dict = get_checkpoint_state_dict(checkpoint_info, timer)
File "C:\stable-diffusion-webui\modules\sd_models.py", line 330, in get_checkpoint_state_dict
res = read_state_dict(checkpoint_info.filename)
File "C:\stable-diffusion-webui\modules\sd_models.py", line 314, in read_state_dict
sd = get_state_dict_from_checkpoint(pl_sd)
File "C:\stable-diffusion-webui\modules\sd_models.py", line 253, in get_state_dict_from_checkpoint
pl_sd = pl_sd.pop("state_dict", pl_sd)
AttributeError: 'NoneType' object has no attribute 'pop'

my bad i'll send screenshots this is a clutter

[strange thunder]

Where should I go to get started with comfy ui? I installed it a couple of months ago, but never actually got around to using it/didn't have the time. "

First and foremost: I have a very precariously set up a1111 and my primary goal is to NOT break it or its extensions. What do I need to do to make sure anything comfy ui related (in terms of install or update) doesn't completely break my a1111?

I also don't want my kohya ss to break. And it seems like both that and a1111 can be casualties of setting up comfy ui. I'd really like to avoid that.

[pure forge]

I feel like it would just be easier to re-set up your computer every time you want to change python versions

[astral comet]

hi guys, so I used DreamBooth to create a model based on some pics, went through the whole tutorial which ended in downloading the created .ckpt file. I put in in the models folder, and when I try to load it it sends me back to the last model used and it says

[frosty portal]

Hello guys, I'm experiencing some weird things with my upscalers, I tried all types of upscalers, different models, amounts, and every time I upscale my model looks deformed, or pixelated, or something like that, I just want it to upscale and specially add detail

Any ideas??

[pure forge]

so now it's way faster but it's just a blur

okay I learned that it's not worth trying to use this on a mac, it's so much simpler on windows

[astral comet]

hi guys, so I used DreamBooth to create a model based on some pics, went through the whole tutorial which ended in downloading the created .ckpt file. I put in in the models folder, and when I try to load it it sends me back to the last model used and it says

[formal trout]

install Comfy, activate the Automatic1111 venv, then run Comfy inside that venv, that won't mess your Automatic1111 installation, and won't take extra space

I did that to save extra space, and I haven't had any issue

then in Comfy you need to install the "manager" (file/extension manager, don't remember), and one by one install all missing nodes, put some models and files in especial folders of Comfy folder for some workflows, some other workflows will use the models in the automatic1111 folders like sd checkpoints and Lora's

[strange thunder]

I got this far and am not sure what to do now.. lol. I actually don't think I mind the extra space it would take up and that's simpler for me than running a venv. Or are you saying that's the only way to do it? Idk if the screenshots are helpful info in this struggle. Dx

I already did a git clone on the manager, but I guess I got ahead of myself a little since the thing itself doesn't want to even launch yet.

[formal trout]

nah you can follow the Comfy installation guide and it will create its own venv and launch thing

[formal trout]

I don't know about that, prob some other user here would know. You can start over again and check if you followed all the installation guide (things like python version, etc)

[strange thunder]

Will updating python or reinstalling xformers risk breaking a1111?

[formal trout]

haha I used to be worry about that kinds of things also, and asked that question. Just check what python version it is needed for Comfy, check your current version. in Windows there is some "add path" thing in the installation of python that was important at least in the past. I don't think it will mess the Automatic installation, but I wouldn't put my hands in the fire. Also I think you can have different versions of python installed and the each venv will use a different version, afaik.

Worst case scenario reinstall Automatic1111, first rename the directory as "stablediffusionOLD", then copy your launch file and models in the corresponding directories. It is a good practice to do that, I've done that when I think my installation was messed up.

[strange thunder]

https://martinfritz.medium.com/work-with-multiple-versions-of-python-on-windows-10-eed1e5f52f07 I think this might be what you're talking about? Where it looks like they tell you to add it to paths and then how to I guess activate the specific version you need?

[formal trout]

yes that's that, Add python x.x to Path

[strange thunder]

Ok thanks. You've at least given me some direction. lol. Although I'm sure I'll make that change and if I'm lucky, it keeps a1111 unbroken, but the comfy ui bat still will have those other issues.. Dx

[formal trout]

no problem, and keep asking here so the guys that know more show xD

[glad fiber]

how can I train a lora on a GTX 1650 (4GB)? I dont care if it takes hours to train, but I want to train on my own computer

it can be textual inversion too, idc

[rotund lava]

you cant train loras with that gpu and textual eats like 6-8gb or vram you could try to train textual inversion with cpu but you would have to search online how to do it

[glad fiber]

isnt there a way where I can combine the power of my gpu and my cpu?

[rotund lava]

no,it can only use one at a time

[glad fiber]

:C

[rotund lava]

someone managed to make it work with a 4gb ampere gpu with kohya-ss and sd-scripts from here https://github.com/kohya-ss/sd-scriptson by using bitsandbytes and xformers but it was an ampere gpu not a turing like yours but you could give it a try

[glad fiber]

thanks, I will see it later

[glad fiber]

link is giving 404

https://github.com/kohya-ss/sd-scripts right link

[rotund lava]

yes that one

you would have to search on youtube tutorials on how to use kohya-ss then u install those scripts to use it together with kohya

[calm cairn]

alright question probably easy one (I don't wanna move all the files to test)
if I have Stable Diffusion on my C drive, can I move the models and loras (and previews) to my E drive and it still works? Or does the program specifically need the folder to be inside its file structure like that?

Also, can I move the whole thing to the E drive and run it from there?

TIA

[simple tartan]

I run all my AI stuff from a different drive

4TB nvme

OS is installed on the 2TB gen5

just move the whole folder (after you close everything)

I'm working on training my LORA again, task from last night failed.... will post the error if it bombs again, currently at 28%....40s/it

[simple tartan]

91%

ModuleNotFoundError: No module named 'bitsandbytes.cuda_setup.paths'
since i'm on AMD, it should be using only CPU, but seems its still attempting an nvidia.

also in the same error log
ImportError: No bitsandbytes / bitsandbytesがインストールされていないようです

going to try as a 1.5 instead of an SDXL for now, while i wait for the brains to come online

[simple tartan]

same error with 1.5

going to force install 0.41.1 and try yet again

[ornate elk]

There is currently no way to train with AMD on Windows that I know of

[simple tartan]

well that sucks, lol

so i need to hurry up and decide which green gpu to add to the system

[ornate elk]

12gb at least. 16gb or more if you want to train sdxl

[simple tartan]

i'm stuck between a A400/4500 or a 4090

so my options are all 20-24GB, lol

[ornate elk]

You can move the models and loras And then add --ckpt-dir "D:/Path/to/Models" and --lora-dir ..
To the webui-user.bat

[ornate elk]

Can you show an example with the txt2img settings?

[simple tartan]

can any of these "apps" SD, Textgen, etc, benefit from the dedicated workstation style cards, or am I wasting money?

[vague compass]

How can I use SD Auto1111 in production. I've 10k+ user requests. How can i handle scaling?

[karmic crown]

exFAT should be used for external drives if you want to be able to use the same drive on both PC and Mac without 3rd party software on Mac - and that's it - in short: just use NTFS for all Windows internal drives

[karmic crown]

show me what you have at the bottom of your UI, something like this:
version: v1.7.0  •  python: 3.10.13  •  torch: 2.1.2  •  xformers: N/A  •  gradio: 3.41.2  •  checkpoint: a88bfb6ffa

I can help you fix your Mac installation, and turbo will work, but it will still be much slower than on your PC

[karmic crown]

for comparison, this is a result on my iMac (Intel) with Radeon Pro 580X 8 GB
You can reduce the time by using Euler A instead of DPM++ 2M Karras
i will try on M1 Macbook Pro later, and post the results from that machine

[astral comet]

hi guys, so I used DreamBooth to create a model based on some pics, went through the whole tutorial which ended in downloading the created .ckpt file. I put in in the models folder, and when I try to load it it sends me back to the last model used and it says

[pure forge]

alright I got a similar result except with too many arms which I posted in #1178805966335119472 but it took 55 seconds

Euler a only reduced it by 4 seconds

[neon vale]

I have an 11GB 1080ti, but when I try to run SDXL with xformers it's excruciatingly slow, even with medvram. no xformers runs fine on high vram

am i lacking ram? is 16gb enough?

[steady thistle]

@ornate elk Hey there! Thanks for all your help recently, it got me back up and running. Do you have a list of useful extensions? It can be difficult knowing what's out there.

[neon vale]

these are my top 3

[steady thistle]

I've heard of adetailer but never tried it. I'll check out the other two, thanks!

[ornate elk]

hey, no problem 🙂 Beside "Tiled Diffusion" whats needed for AMD, i can recommend "ControlNet" (its the most feature rich extension out there and very powerfull),
also for prompting the extension called "Booru tag autocompletion" is very helpfull. "Adetailer" is okay too for face and hand fixes. For FaceSwap "Reactor" is great, but can also be done with ControlNet.
There is also the "lobe Theme" Extension for a nice looking UI.
And then it depends on what you want to do. There are tons of Extensions for example for video stuff etc.

[steady thistle]

I use ControlNet a lot and ReActor all the time, but I'm having trouble with Canny and Depth in SDXL. Is there a different version I need? I'm hesitant to install stuff if I'm not sure about it after I broke my install last time. lol

[ornate elk]

here oyu find all sdxl controlnet models:
https://huggingface.co/lllyasviel/sd_control_collection/tree/main

[steady thistle]

Thanks!

[ornate elk]

maybe use sai or t2i adapter canny version

[steady thistle]

Ooh, okay, now it makes sense.

I immediately saw what I had been missing.

Could I suggest that maybe there be a channel just for extensions? There are so. many. extensions out there.

[hot spindle]

@karmic crown 👋 Thanks for the help sorting my batch issues on github , its usable now, not the way I tought it should be,but hey..it works

[ornate elk]

i think hat wouldnt help as then everyone asks about their issues there, like they do here rn.
Because there are so many extensions there wouldnt be a big change to here.
we have #📝｜prompting-help where also everything extension related can be ask for better outputs and here if its more on the technical site.

[steady thistle]

That's fair

I'll check that out more.

[ornate elk]

sure, and we have #1072238304042438758 where also every sd thing can be discussed

[steady thistle]

Is there a good extension for prompt history?

[ornate elk]

yes there is this:
https://github.com/ArtVentureX/sd-webui-agent-scheduler

[uncut jewel]

Hey guys! I updated comfyUI and when i started i got a grey screen that i cant do much.. So i tried to load a workflow and this is the error message i get

[calm cairn]

hey sorry I fell asleep--did you mean to add these after "set COMMANDLINE_ARGS="?

[ornate elk]

yes

[calm cairn]

cool ty very much, cant wait to free up a few hundred gigs on my solid state

[uncut jewel]

Can anyone help me regarding this issue pls ? 🥹 🙏

[calm cairn]

do you know if spaces are allowed (ART/Stable Diffusion/models as opposed to ART/Stable_Diffusion/models)?

[neon vale]

yeah, just put it between " "

[astral comet]

@ornate elk hi guys, so I used DreamBooth to create a model based on some pics, went through the whole tutorial which ended in downloading the created .ckpt file. I put in in the models folder, and when I try to load it it sends me back to the last model used and it says this

[vocal burrow]

in theory yes, but just to be safe and because not every extensions is built equally, I d say it s best to avoid weird characters, spaces, etc

[ornate elk]

hey, what was your gpu again?

[astral comet]

@ornate elk 1070

[karmic crown]

I just generated about 20 images (with spaceman as a prompt), and only one had extra arms. A good negative prompt or text inversion should help, but unfortunately, some seeds might generate deformed images even with a good negative prompt or text inversion (like 3090860638, for example). I tried a few different negatives, but none helped

your speed is too low. what do you currently have as your command line parameters? please try those two:
--skip-torch-cuda-test --opt-sub-quad-attention --upcast-sampling --no-half-vae --medvram-sdxl
--skip-torch-cuda-test --opt-sub-quad-attention --upcast-sampling --no-half-vae

if it still takes more than 30 seconds, try to disable acceleration in your browser since i experienced cases when the browser took more than 35% of my GPU (check activity monitor)

[karmic crown]

hello, please try those two command line params (and let me know if there is any difference in speed or usability of controlnet):
--skip-torch-cuda-test --opt-sub-quad-attention --upcast-sampling --no-half-vae --medvram-sdxl
--skip-torch-cuda-test --opt-sub-quad-attention --upcast-sampling --no-half-vae

also, check if your browser takes too much of your GPU (check the activity monitor). mine was used more than 35% in some cases, if your browser takes too much of your GPU, disable acceleration in browser

[ornate elk]

okay, i didnt used dreambooth before, but while training a lora i got the same issue. (its GTX card related). make sure the model is 1.5 based and the sizes of the images are mostly 512x512. Also if there is any Setting for bf or fp set it to fp16. Also Uncheck AdamW8bit or (activated it AdamW) if possible.
Also i would recommend using khoya_ss tool for training Loras or Checkpoints. Its a seperate tool that wont break you SD webui like Dreambooth will. It also has more Settings.

[karmic crown]

i will be back in a few hours, buy for now

[hot spindle]

will do, tks, and will answer back

[karmic crown]

@pure forge please check https://github.com/AUTOMATIC1111/stable-diffusion-webui/discussions/5461#discussioncomment-8050803 so you can see what parameters are really required for Macs and which are "optional", so you can experiment yourself

@hot spindle since that was my anser for you, you already saw that 🙂

[hot spindle]

I always do my bechmarch using benchmark option in system info extension - with SDXL model loaded wich is the one I usually use - do you think its a fair assesment or should I use img2img (my most typical scenario)?

[astral comet]

@ornate elk ty! Ill try that out

[pure forge]

with the first set of parameters it took 26 seconds

[karmic crown]

I would recommend testing with real image generation. When I am doing the test, I usually generate 4 images (4 batches of 1 image - that works faster and uses fewer resources than 1 batch of 4 images) and see how long that takes. When you generate only 1 image, timings can be a little inaccurate. I noticed that in some cases, the last step takes too much time compared to other steps

i have almost never noticed that slowdown at the last step with Euler A, but I have noticed that from time to time with DPM++ 2M Karras and some other samplers

[pure forge]

my browser used .1% GPU but Python used 85%

[hot spindle]

ok,tks , like I said before, for some reason img2img on my rig crashes when doing more than 1 image batch , but it works ok if I generate them at same time

[karmic crown]

my m1 and my iMac are much slower than your Macs, so you should get better results than me

[karmic crown]

that should not happen, i will get back to you later, i am late for a meeting

[hot spindle]

Batch count more than 1 - crash current generation (but not A1111) , Batch size is ok with 4

ok,sure, dont wanna take your time, will test all that and report back with findings

[karmic crown]

batch count = 4 + batch size = 1 should work faster and use less resources than batch count = 1 + batch size = 4

[hot spindle]

yes, I realize that, no idea why..it has allways been so. Tks and talk later

[calm cairn]

So, I have only succeeded at moving the LoRA directory. For some reason, the checkpoints have remained at the old location. Here's the argument I typed in, in full context:
set COMMANDLINE_ARGS=--xformers --medvram --no-half --ckpt-dir "E:/Life/Projects/ART/Stable_Diffusion/Models" --lora-dir "E:/Life/Projects/ART/Stable_Diffusion/LoRA"

I know word wrap makes that look like a mess, but I'm hoping you can tell me why one would work and the other fail--I got no error messages in the launch.

(the lora directory is empty and the models directory has exactly one checkpoint with preview image)

[ornate elk]

so all the other models are now moved but dont show up?

what is the one model in the models folder? sdxl or 1.5 based?

[calm cairn]

No, the models are still on the C: drive in the old location, and the new folder on E drive is empty. I have redirected SD to the new folder, but it still shows me the models on C drive.
The model in the folder on E is "anylora", a 1.5 model

The clients display for Loras tells me that the directory is empty, and if I'm successful with moving, then it should also tell me that (rather than a lot of models), there is only 1 model

[ornate elk]

so loras are now on E and also shown in the webui?

can you try end the path with /" ?

[calm cairn]

trying that now

let me be a bit more clear---before today, all models + loras on C drive
I want to go to E drive (get it off my SSD)
so I've made 2 empty folders on E: and I want SD to understand that before I move a lot of files
Right now only 1 ckpt and 0 loras are on E because I want it to work before I swap, make sense?

this is the message for loras

(success on that one)

but models is showing my old list of models on C, rather than just the 1 on E (which would be correct)

added slashes at end, no change

I am sure the machine is updating the list, because the one model I moved over (anylora ckpt) is no longer in the list