🤝｜tech-support

[vocal burrow]

it's for developers.

[sweet zenith]

Is this the correct channel for API questions?

I'm struggling to get the sd3 endpoint to work with img-to-img, text to image works fine

async function imgToImg() {
  const imageData = fs.readFileSync(
    'src/stability-ai/src-imgs/pink_phones.png',
    'utf8',
  );
  const buffer = Buffer.from(imageData, 'utf8');
  const formData = new FormData();

  formData.append('image', buffer);
  formData.append('mode', 'image-to-image');
  formData.append('model', 'sd3');
  formData.append('strength', 0.35);
  formData.append('prompt', 'apples');

  console.log(formData);
  const header = {
    headers: {
      ...formData.getHeaders(),
      Accept: 'image/*',
      Authorization: `Bearer ${apiKey}`,
    },
    responseType: 'arraybuffer',
    validateStatus: undefined,
  };
  const url = `https://api.stability.ai/v2beta/stable-image/generate/sd3`;

  const response = await axios.post(url, formData, header as any);

  if (response.status !== 200) {
    console.log(response.status);
    console.log(response.data.toString());
    throw new Error(`Non-200 response: ${await response.statusText}`);
  }

  if (response.status === 200) {
    fs.writeFileSync(
      'src/stability-ai/out-imgs/v2_img2img.png',
      Buffer.from(response.data),
    );
    console.log(response.headers.seed);
  }
}

I keep getting the error:

{"name":"bad_request","errors":["image: is required"]}

[ornate elk]

Hey, zluda can work on Linux too, but is best used on windows.
Because on Linux AMD cards can use Rocm natively which is even faster than zluda

Which model did you tried to use?

From which webui is the error?

[sweet zenith]

Has anyone had luck using the sd3 img-to-img endpoint?

(using typescript or js)

[ornate elk]

can you try to install auto1111 ?

because sd.next uses a different backend, thats why you get the diffusers error

you need to have rocm installed

[true bridge]

can anyone help me setting up sdxl temporalnet in comfyui?

[ornate elk]

since Pop OS is based on ubuntu you can try this guide:
https://www.youtube.com/watch?v=NKR_1TUO6go&themeRefresh=1

oh this error appears for many users the last two days

i still dont know whats causing it

try load an other model

for example dreamshaper v8

what does the cmd shows then?

try add --medvram-sdxl --upcast-sampling at the COMMANDLINE_ARGS=

you could add them after the webui.sh

as launch params

[sweet zenith]

reading in the image using fs.createReadStream caused this to work.

  const imageData = fs.createReadStream(
    'src/stability-ai/src-imgs/pink_phones.png',
  );

[ornate elk]

nope, hes refering to his own question a bit above

[ornate elk]

hey, can you show the whole cmd screenshot?

maybe its trying to use your igpu if you have an amd 7000 cpu or intel

[ornate elk]

what model is loaded? 1.5 or sdxl?
also disable FILM interpolation

what model did you loaded at the top?

[sweet zenith]

what format is the image supposed to be sent to for sd3 image to image API?

base64 encoded string?

[sweet zenith]

Hello -- do you know if there is a place where people are talking about using the API?

[ornate elk]

Maybe in #1003207327203209236

[rocky adder]

Thanks I solved it

[ornate elk]

Nice, what was the cause?

[thick gust]

Where could I find support, and more information about having system like these properly and professionally configured on one of my machines?
Of course, I'm referring to a paid service, for company that I work for. Thanks

[trail sigil]

does anyone know a fix to this?

[unique bluff]

I managed to add a launch arg (with the help of ai) and used the command to launch using cpu only (safe to say it’s a much smaller toll on my computer lol)

But other than that didn’t manage much else on that code

[ripe bronze]

Hi, I am very new to compyui. I have downloaded some popular workflows from websites, downdloaded them, and used Manager to install missing nodes. Despite doing with all of the workflows, whenever I try to "queue prompt" none of the workflow generate an image because there's always a red error message with multiple various lines.

My question is: when I see these messages I don't know how to resolve the issues?

I'll post one example below from a fairly simple workflow. Not sure how to resolve / fix

Completely lost on how to fix any of these messages

[fair oxide]

looks like its asking for missing checkpoints, vae, and loras

all of which should be available for download at official civitai.com website

[unique bluff]

What kind of variables should I set on my webuiuser.sh?

To make sure the lowvram and everything else is correct

[rocky adder]

Stable Diffusion failing again and again... When try to use AnimateDiff ... I just press Generate, and it does nothing... it goes back to its initial state. I have re started my computer several times and ... well, it was working for many hours today...
How is it so unreliable?? Any tips on how to make the use of Animate Diff more reliable and... Stable??
Thanks a lot

Now it says "In queue 1/1", but when you look in Task manager, there's actually nothing going on

It's just generting an image now! And it cannot even handle it...

Please somebody give me a tip, this is so weird... Don't know what to do

[unique bluff]

Okay, finally it worked

[rocky adder]

I got it! It' s when I paste my prompt... then AnimateDiff fails to work. Can somebody tell me how to properlly paste a prompt in txttoImage when using AnimateDiff?? It's really a pain to write it everytime.

[cursive needle]

Why is this happening?

I can't find any settings to fix this. I've tried Hi res on and off. Different levels of noise, most of the sampling methods. It generates fine until the end, then the images always turn into this noisey mess

[noble belfry]

disable the hires. or at the very least turn its denoise strength down to ~0.3

also is that an 8 step version of dreamshaper?

[cursive needle]

Its the newest one

seems to be the only model i am having issues with. I'll see if another one does better

[noble belfry]

from your cfg levels those look like LCM parameters. which needed a specific lcm sampler

[vast mica]

Ok so, to preface, ive completely reinstalled SD 1.5 onto my PC
reinstalled python version 3.10.11
this is currently my Webui-user notepad set COMMANDLINE_ARGS=--use-directml --opt-sub-quad-attention --no-half --disable-nan-check --autolaunch --xformers
im trying to use an AMD GPU (8GB vram) to run SD
i am using the proper Automatic1111 for AMD
these are the errors im getting
while i am getting errors, it does properly open up for me to try and SD but when i try with my model, it loads to 100% and then fails

this is the error when i try to run a prompt

[rocky adder]

How does this setting "picts to interpolate" work? I don't know where to press to start.. is in the right bottom corner
Thanks!

[sweet zenith]

Thanks

[crude gorge]

Hello everyone i tryed to look for a similar questions but didn't have any luck so i'm asking you guys for help. My Automatic1111 is taking long pause between image generation
When i generate a batch of 10 images (1024*1024) Automatic1111 generate the image in like 2-3 seconds but then before going to the next image generation in the batch it pauses for almost 1 minute any idea what is causing this ? maybe an options or extension wrongly configured ? (no error message showing in the log)
thanks for your help 🙂

Additionnal informations :
the launching arguments : --xformers --no-half-vae --api
and while the VRAM should not be a problem in my config (24G) the hard RAM maybe is since i have "only" 32 G RAM in DDR4

Thanks in advance for your precious help !

[ornate elk]

Could be the wrong VAE that is used. Check in the settings which VAE is selected

[ornate elk]

Are you using Batch Size or Batch Count?

Sdxl models or 1.5?

[crude gorge]

batch count (so image are made one by one not multiple at once) SDXL model (when i make 1.5 512x512 i don't have this problem)

[crude gorge]

as for the VAE it's in auto

[ripe bronze]

For Comfui, do you know whether running a AMD cpu and gpu vs. NVIDIA cpu and gpu of very similar price range woul would generate images at a similar speed or not?

Someone in a video claimed that the not using Nvidia gpu for comfyui would run WAY closer. Trying to see if anyone knows how true that would be?

[serene dirge]

*** Error loading script: reactor_swapper.py
*** Error loading script: reactor_faceswap.py
*** Error loading script: reactor_api.py

Hey guys I've been dealing with this issue for hours now, I've tried everything, Installed BOTH VS2022 and VS C++ Build tools (Along with Desktop C++, Python, and Extension), python 3.10.6, activated venv and pip insightface, placed inswapper_128.onnx in insightface folder and yet ReActor extension isn't still showing up and I'm on my wits end.

[crude gorge]

if you want to see the full list of available VAE

[ornate elk]

Try remove the --no-half-vae from your webui-user.bat
Or get the fp16 sdxl VAE

[crude gorge]

damn smart

[ornate elk]

Depends on the card. But for ai stuff nvidia is better right now.

[crude gorge]

is this one https://huggingface.co/madebyollin/sdxl-vae-fp16-fix/tree/main ok ? the name does not refer to fp16 but in the description it looks like it's the right one will 1111 be able to pick the correct one still ? or it's based on the name ?

[ornate elk]

What does the cmd shows when starting the webui-user.bat?

[ornate elk]

Yes this one

Then don't select automatic

It won't pick up any vae

[serene dirge]

Sorry if I had to snapshot it due to character limit, that's the initial start the rest are the following *** Error loading script

I also use web-ui 1.7 due to mov2mov isn't working on the latest update if that matters

[crude gorge]

thx will try that 🙂 ❤️

[serene dirge]

File "C:\stable-diffusion-webui-1.7.0\venv\lib\site-packages\albumentations\augmentations\geometric\resize.py", line 6, in <module> from pydantic import Field, ValidationInfo, field_validator ImportError: cannot import name 'ValidationInfo' from 'pydantic' (C:\stable-diffusion-webui-1.7.0\venv\lib\site-packages\pydantic\__init__.cp310-win_amd64.pyd)

There's also this right after the last screenshot mb

[ornate elk]

@trail sigil @serene dirge Here is a temporary fix for that known issue:
https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/15551#issuecomment-2061997952

@serene dirge and here is the fix for mov2mov for webui version 1.9.0:
https://github.com/Scholar01/sd-webui-mov2mov/issues/146#issuecomment-2054157728

[serene dirge]

Omg you're a God, I wished I would've asked way sooner here tysm! I'll check the move2mov for the latest version too

[quaint wave]

Hello, when I'm trying to generate my images with SDXL and HiRes. Fix it throws the error
RuntimeError: The model is mixed with different device type The model is mixed with different device type
It can generate without hires fix though. I'm using SD Forge and A750 with ipex, so it might be happening because of it. Any ideas on how to fix it?

[ornate elk]

Hey, what's your GPU?
The first issue is that you have --disable-nan-check and --xformers in the webui-user.bat.
Xformers is only for nvidia cards.
The second message isnt an issue and can be ignored.
But if you have an 5700xt or newer you should use my newer install guide for the webui using zluda which is much more faster and less vram hungry than directml.

[trail sigil]

that works, you're a miracle

[vast mica]

aha, i have a 8192MB ATI Radeon RX 570 Series (MSI) 💀
and what i ended up doing what completely deleting the Venv folder to restart
and set my ARGS to
set COMMANDLINE_ARGS=--use-directml --opt-sub-quad-attention --no-half --disable-nan-check --autolaunch

so it works with what i have now, but im not sure if its using my GPU or not

[ornate elk]

Remove --disable-nan-check
It doesn't to anything other than making images black

[vast mica]

got it

[ornate elk]

Check the vram usage in taskmanager when generating

[vast mica]

im not sure why i never decided to check TM to see if GPU was being used or not in which case it is in fact being used thank you

[ornate elk]

Never seen anyone using forge and Intel arc together. But you could try setup SDnext with ipex and test it there. It should have a better Intel support.
For forge try using a lower upscale value

[quaint wave]

I have tried sdnext

It works and doesnt

its random lol, one generation it will work, the second it throws RuntimeError: Native API failed. Native API returns: -999 (Unknown PI error) -999 (Unknown PI error)

forge has been the best one so far but it has some issues as well

openvino was the worst lol.

[ornate elk]

Hmm okay.
What you can try in forge is to go into settings and there set the RNG to CPU

[quaint wave]

I already have it like that

[ornate elk]

Okay good

[quaint wave]

I also tried NV, similar results

only one that doesnt work is GPU

[ornate elk]

How much RAM (not vram) do you have?

[quaint wave]

8

oh ram

32

[ornate elk]

Okay 32 should be enough

But check your windows pagefile settings. That its set to C drive only and on automatic

And deactivated for any other drive

Also make sure to have 15gb free space on C

But if the GPU only has 8gb of vram thats to less for upscaling sdxl images.

[quaint wave]

I've got 3 drives but only C is active

I dont think its a system failure rather its an ARC issue or webui

[ornate elk]

Yea I think so too. Can you check the vram usage in task manager when trying to upscale sdxl?

[quaint wave]

https://waa.ai/cqNp.png

This is where it changes to upscale

Cant even begin to try it lol

[ornate elk]

Okay

What you can instead use for upscaling is also the SD upscale script in img2img.

Maybe that won't get that error

For me it seems webui forge tries to use CPU for specific tasks when it can't use the GPU

[ornate elk]

Oh have you tried upscaling 1.5 models?

[quaint wave]

Yes they work and they dont lol

50/50

[ornate elk]

Same error?

[quaint wave]

Not really

couple of different errors

[quaint wave]

RuntimeError: Native API failed. Native API returns: -997 (Command failed to enqueue/execute) -997 (Command failed to enqueue/execute)

[ornate elk]

Okay no clue on that one

I see some issues on forge github where other Intel ARC users have a similar error

Here is a thread:
https://github.com/lllyasviel/stable-diffusion-webui-forge/discussions/455

[quaint wave]

yes I have comments in there lol

changing steps, upscaler, resolution doesnt have any effect

[ornate elk]

Okay can you try the SD upscale script?

In img2img

[quaint wave]

this is what it gave me

lol

[ornate elk]

Ohh

[quaint wave]

any other webui's i can try for ipex?

[ornate elk]

The only other one would be Shark webui

[quaint wave]

https://github.com/nod-ai/SHARK

this one?

[ornate elk]

Yep

Make sure you put the .exe into a new folder

Before running it

[fair oxide]

Who gonna help me with api interaction

[quaint wave]

SystemExit: Could not compile Unet. Please create an issue with the detailed log at https://github.com/nod-ai/SHARK/issues

This is what I get when trying to generate now in Shark

im lost honestly lol.

[primal glen]

I just installed Stability Matrix and it's giving me this error message when I try to launch Stable Diffusion WebUI:

Traceback (most recent call last):
File "C:\Users\Master\AppData\Roaming\StabilityMatrix\Packages\stable-diffusion-webui-forge\launch.py", line 1, in <module>
from modules import launch_utils
ModuleNotFoundError: No module named 'modules'

[ornate elk]

What's your GPU?

[rotund musk]

Whats the best non banned colab for stable diffusion btw

Is it the interactive one or is there a better one? Bc i heard the interactive one has issues

[primal glen]

what do modules have to do with my GPU?

[remote ivy]

Anyone else find that sdxl lightning models produce grainy outputs?

[ornate elk]

Just a basic question because Forge works not that good with AMD GPUs.

Also if you have an nvidia GPU with more than 6gb vram, I would recommend you to install automatic1111 instead of Forge

[noble belfry]

well theyre not going to produce top tier fidelity given their nature. and given their low steps few extensions work properly as 4-10 steps isn't enough room render time to their models. you can absolutely beef them up with an upscale though, even an extra 2 steps at 1.25 scale does wonders on a med-low denoise. also had some good results with PAG but its almost like using a ouija board trying to get the parameters right for lightning models

[primal glen]

let's get SD running at all before trying to get it to run well. How do I tell the program where the modules are?

[ornate elk]

Modules is something that should be installed by the webui itself

Mostly a wrong python version can cause such issue where the modules can't be found

[gray falcon]

Hello guys, is there any mod that is responsible for Stability API? I need to get some information about our usage history.

[karmic crown]

I was on a business trip, and I had a lot of work before and after returning 🤷🏻‍♂️

yes, I have seen errors like that, but not inside the SD

something was wrong with SSL cert on the server, or at least cURL thought it was an error

@quiet swift have you fixed your error?

[primal glen]

Python is 3.12.0, pip is 24.0. I just ran pip's updater to make sure everything was up to date.

[ornate elk]

Python 3.12 is not supported by forge or auto1111
you would need 3.10.11 64bit

[vocal burrow]

np

[vocal burrow]

and don t forget to delete the venv directory (if you re using one with forge)

[karmic crown]

i tested 3.11 and worked fine for me, but i also advice sticking with version recommended by developers

[primal glen]

cool, okay. I've had this problem before and ended up with a mess of different versions of python on one computer. How do I change to that version without leaving traces of 3.12 behind?

[primal glen]

I saw this on a reddit thread, I deleted venv and reinstalled it. Should I delete it and keep it deleted?

[vocal burrow]

you simply uninstall the others python like any other programs

(assuming you don t need any other python version)

[vocal burrow]

reinstalled it ??

what do you mean ?

[fair oxide]

@ornate elk any public access to sd3 yet?

[vocal burrow]

through the API, that s all

[fair oxide]

yeah but that requires sub?

[vocal burrow]

I think it requires coins/credits. not necessarily a sub.

So you can use your free credits and then buy a bunch more if needed.

[vast crest]

hi, I still have this problem that after creating a sequence: loopback-wave I click on the icon as in the photo you see, which should open the folder

instead it opens the folder:img2img-grid. How do you resolve this error?

[fair oxide]

ah ok, i'm not sure tho how acquiring coin/credit system works

[primal glen]

I can't find the page I found this on, but it said step #1 is to delete the folder called 'venv', then steps #2-5 were running commands that put the 'venv' folder back but in the supposedly proper place. I can try deleting it altogether to see if that does anything

[vocal burrow]

that sounds weird.... the venv will recreate itself when launched. No need to run anything to reconstruct it.

[primal glen]

yep. I deleted it, looks like it's recreating like you said.

[vocal burrow]

cf https://platform.stability.ai/pricing

[primal glen]

okay, I deleted venv. Now I'm getting this error: https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/1742

[gritty ridge]

How is inpaint upload different from inpaint sketch- I mean what is the format etc… could it be technically precise mask

[vocal burrow]

what s your setup/hardware ? what s the full log ? Copy paste it (the whole thing) into a file and drop said file in here.

[karmic crown]

https://dev-discuss.pytorch.org/t/pytorch-macos-x86-builds-deprecation-starting-january-2024/1690

Bye Bye Intel Macs 👋

[vocal burrow]

ooooh

I can see from here the incoming wave of "why is it broken" mac users

how to fix it ? pay the apple tax / buy a new machine

And with sd3 incoming A1111 might hit that wall relatively soon.

wait and see

[bitter zenith]

that’s not apples fault lol, just keep your old Mac

[fair oxide]

what do you mean a1111 will hit the wall?

[vocal burrow]

Old mac users will have errors when A1111 update pytorch to some versions that don t support intel mac (unless they do it properly and set a lower pytorch version requirement for them... but it will probably be hard to maintain on the long run)

[fair oxide]

oh that, well i was wondering if a1111 would be incompatible with sd3, ... comfyui is not a viable choice for many for several reasons, dont wanna spend 15 mins setting things up to write a 15 second prompt, but anyways im good, not a mac user 🙂

[vocal burrow]

I don t have any news about A1111 regarding sd3 support. It will probably work but it won't be an easy task to support it.

So I don't know, it depends if they had early access to it, if they had time, if if if if

[fair oxide]

a1111 might take days to catch up to all the features introduced in sd3, but i know users didnt fall short relying on a1111 when sdxl first came out

[karmic crown]

i will change my script to use torch 2.1.2 for intel and the latest stable for arm
but i am waiting for 2.3.0 to be released first, since 2.2.0 does not work properly on new macos

just tell them to use my script 🙃

[rocky adder]

Can anybody guide me a bit? How to use this Pics to Interpolate feature inside Deforum?
https://media.discordapp.net/attachments/1002602742667280404/1230727876098396241/how_to.PNG?ex=66233c27&is=6621eaa7&hm=cd8ce77fc6bbba580d6a81d1961e905646f9ae0032e5b68b660fdc550af9991b&=&format=webp&quality=lossless&width=1479&height=1330

[vocal burrow]

let s just hope that A1111 won t be using <= 2.3.0 exclusive feature in the future

[karmic crown]

i will just print a message "please buy new mac" in that case 🤣
i am bad

[rocky adder]

What is the difference between AnimateDiff and Deforum?

[vocal burrow]

They work differently, different settings, different outputs, etc... For deforum support you ll probably have better chances asking for help in their discord.

[primal glen]

Version: f0.0.17v1.8.0rc-latest-276-g29be1da7
Commit hash: 29be1da7cf2b5dccfc70fbdd33eb35c56a31ffb7
Traceback (most recent call last):
  File "C:\Users\Master\AppData\Roaming\StabilityMatrix\Packages\stable-diffusion-webui-forge\launch.py", line 51, in <module>
    main()
  File "C:\Users\Master\AppData\Roaming\StabilityMatrix\Packages\stable-diffusion-webui-forge\launch.py", line 39, in main
    prepare_environment()
  File "C:\Users\Master\AppData\Roaming\StabilityMatrix\Packages\stable-diffusion-webui-forge\modules\launch_utils.py", line 431, in prepare_environment
    raise RuntimeError(
RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check```

where is "commandline_args"?

[vocal burrow]

and what s your setup / hardware ?

[vocal burrow]

it should be in C:\Users\Master\AppData\Roaming\StabilityMatrix\Packages\stable-diffusion-webui-forge\webui-user.bat. But because you re using StabilityMatrix they might also be stored somewhere else on top of that, never used it so I dont know.

[primal glen]

For "setup/hardware" you want specs about the PC I'm running stabilitymatrix on, right? What information do you need besides that I'm running windows 10?

[vocal burrow]

cpu, gpu, ram

[primal glen]

Device name HAL-9000
Processor Intel(R) Core(TM) i7-6600U CPU @ 2.60GHz 2.81 GHz
Installed RAM 32.0 GB (31.9 GB usable)
Device ID 4BB2FF74-11EA-4277-9C7B-088C27347D60
Product ID 00330-51681-05322-AAOEM
System type 64-bit operating system, x64-based processor
Pen and touch No pen or touch input is available for this display

GPU

[vocal burrow]

ok so you indeed do not have a proper gpu.... You can set the --skip-torch-cuda-test it s telling you to. But doing that will make SDwebui run on "cpu only mode" aka extra super duper slow

[primal glen]

yeah it's fine if it takes ages to do anything. I'll worry about improving performance once there's actually performance to improve.

[primal glen]

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--skip-torch-cuda-test

@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

the contents of webui-user.bat

I'm still getting the same error

[vocal burrow]

probably because stabilitymatrix don t use webui-user.bat and call webui.bat directly with its own arguments

yup indeed

[primal glen]

so I need to add "--skip-torch-cuda-test" somewhere else to get stabilitymatrix to use it

[vocal burrow]

yup somewhere that looks like that

I assume there will be a "skip torch cuda test" checkbox or something like it

I love it when wrappers go out of their way to make our life """"easier""""

yeah looks like there should be such an option https://github.com/LykosAI/StabilityMatrix/blob/c18e93271e8a4d312e1fcdef2a8c51ddfca63974/StabilityMatrix.Core/Models/Packages/SDWebForge.cs#L94

[primal glen]

of course! why would I want to go through the trouble of modifying .bat files directly when I can sift through some arcane UI instead?

it's installing now. I'll come back when it throws the next error message.

[vocal burrow]

I mean tbf you could have choose to install forge directly too instead of going through StabilityMatrix :p

That s part of the traps you can t predict until you set them off

hopefuly it should be fine from here

[primal glen]

yeah I'll do that in the future but I'm not going to uninstall it right when I finally get it working

[vocal burrow]

fair enough

[primal glen]

new error message is too long to paste in discord so I've pasted it here: https://pastebin.com/ETaUNPAF

[karmic crown]

just drag & drop file here

[primal glen]

I need to tell the program not to look for an NVidia driver. How do I do that?

[karmic crown]

try --always-cpu

that should help with forge without nvidia

but it will be PAINFULLY slow

or even better try --disable-xformers first

[primal glen]

this is the error I get when I check the "always-cpu" box

[karmic crown]

--skip-torch-cuda-test does not work for forge as far as i remember

[vocal burrow]

should be no need for it as they did nt enable it.

[primal glen]

there's no "disable-xformers" box

[vocal burrow]

but yes, use --always-cpu

[karmic crown]

do you have some box where you can enter whatever you want?

i have never used matrix

[primal glen]

"extra launch arguments" has a text box I can type stuff in. I'll type "--disable-xformers" into that.

[karmic crown]

looks like right place

[primal glen]

I get this error message again

[karmic crown]

my advice is to go to https://github.com/lllyasviel/stable-diffusion-webui-forge and than download forge one-click pacakge >>> Click Here to Download One-Click Package<<<

just put it in some other folder

it should work without any modifications, you just might need to add --always-cpu

[primal glen]

if I didn't have stuff already installed on my computer I would do that. Can you promise that the version of stablediffusion I'm working on installing right now won't get in the way of the install you're recommending?

[vocal burrow]

https://github.com/lllyasviel/stable-diffusion-webui-forge/issues/674

suggested hack should "fix it" enough for now.

Use requirements_versions.txt as the last post suggested

[karmic crown]

pydantic==1.10.11

add that in the end of requirements_versions.txt

[primal glen]

huh, okay, lemme try that.

[karmic crown]

it should ened like this

httpx==0.24.1
basicsr==1.4.2
diffusers==0.25.0

just add that like this and save

httpx==0.24.1
basicsr==1.4.2
diffusers==0.25.0
pydantic==1.10.11```

then try to lauch it again

[primal glen]

adding "pydantic==1.10.11" to the requirements_versions.txt looks like it worked. It's 5% finished generating a test image 🙂

[valid mulch]

When running a batch in Comfy, it appears from the CMD that it's re-loading the model for each gen. This means the first gen is fast and they get slower and slower until eventually it's taking hours to complete a single image. Is there something I could be doing wrong in my workflow that's causing it to want to pull the models over and over?

I see "Loading 1 new model" and "Requested to load SDXL" repeated several times for each new image.

My flow isn't anything complex... one model with a couple of loras followed by face detailer and sd ultimate upscale...

[primal glen]

success! Thanks so much for the help.

[thorny mortar]

hello

[karmic crown]

friendly advice, use some "fast" model, that require less steps to get decent image, like lightning or lcm

[thorny mortar]

I try to run dreambooth to train new model with diffuser packet

My setup code is

cd diffusers
pip install .
cd examples/dreambooth
pip install -r requirements.txt```

It work

Then I run a training code like tutorial

```accelerate launch .\diffusers\examples\dreambooth\train_dreambooth.py ^
--pretrained_model_name_or_path="C:\Users\dungt\Desktop\diffusers\assets\animeScreencap.safetensors" ^
--output_dir="C:\Users\dungt\Desktop\diffusers\assets\model" ^
--instance_data_dir="C:\Users\dungt\Desktop\diffusers\assets\source_img" ^
--instance_prompt="myxmystyle, anime, 1girl"```

But I alway get this error, I dont understand where I was wrong

  File "C:\Users\dungt\AppData\Local\Programs\Python\Python310\lib\site-packages\transformers\utils\hub.py", line 398, in cached_file
    resolved_file = hf_hub_download(
  File "C:\Users\dungt\AppData\Local\Programs\Python\Python310\lib\site-packages\huggingface_hub\utils\_validators.py", line 111, in _inner_fn
    validate_repo_id(arg_value)
  File "C:\Users\dungt\AppData\Local\Programs\Python\Python310\lib\site-packages\huggingface_hub\utils\_validators.py", line 165, in validate_repo_id
    raise HFValidationError(
huggingface_hub.utils._validators.HFValidationError: Repo id must use alphanumeric chars or '-', '_', '.', '--' and '..' are forbidden, '-' and '.' cannot start or end the name, max length is 96: 'C:\Users\dungt\Desktop\diffusers\assets\animeScreencap.safetensors'.

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "C:\Users\dungt\Desktop\diffusers\diffusers\examples\dreambooth\train_dreambooth.py", line 1443, in <module>
    main(args).........```

[karmic crown]

this looks strange

it should be --pretrained_model_name_or_path= i guess

but it might be just a display

[thorny mortar]

just a display

[karmic crown]

try to rename this source_img and to use something like sourceimg, since you have this error
Repo id must use alphanumeric chars or '-', '_', '.', '--' and '..' are forbidden, '-' and '.' cannot start or end the name, max length is 96:

i am just guessing

i have never used that

[thorny mortar]

I just try, same error

[karmic crown]

https://github.com/huggingface/diffusers/issues/736

check if that helps

[thorny mortar]

yes I have read that before

[karmic crown]

https://www.reddit.com/r/StableDiffusion/comments/zmniuo/dreambooth_in_automatic_error/

[thorny mortar]

hmmm

I think you was right

I just change all folder and file in same folder

now my code look like this

[primal glen]

https://civitai.com/models/195519/lcm-lora-weights-stable-diffusion-acceleration-module <--like this?

[thorny mortar]

--pretrained_model_name_or_path="animeScreencap.safetensors" ^
--output_dir="model" ^
--instance_data_dir="sourceimg" ^
--instance_prompt="myxmystyle, anime, 1girl"```

it work with different error

  File "C:\Users\dungt\AppData\Local\Programs\Python\Python310\lib\site-packages\huggingface_hub\utils\_errors.py", line 304, in hf_raise_for_status
    response.raise_for_status()
  File "C:\Users\dungt\AppData\Local\Programs\Python\Python310\lib\site-packages\requests\models.py", line 1021, in raise_for_status
    raise HTTPError(http_error_msg, response=self)
requests.exceptions.HTTPError: 401 Client Error: Unauthorized for url: https://huggingface.co/animeScreencap.safetensors/resolve/main/tokenizer/tokenizer_config.json

The above exception was the direct cause of the following exception:

Traceback (most recent call last):
  File "C:\Users\dungt\AppData\Local\Programs\Python\Python310\lib\site-packages\transformers\utils\hub.py", line 398, in cached_file
    resolved_file = hf_hub_download(
  File "C:\Users\dungt\AppData\Local\Programs\Python\Python310\lib\site-packages\huggingface_hub\utils\_validators.py", line 119, in _inner_fn
    return fn(*args, **kwargs)```

I will continue check that

[karmic crown]

set filter to show sd 1.5 lcm based models

https://civitai.com/models/4384?modelVersionId=252914

you can try this one

[primal glen]

where does that one say "lightning" or "lcm"?

[karmic crown]

be sure to use low CFG, or you will get a messy images
something like 8 steps is enough

[karmic crown]

those are SDXL, but that will probably be too slow on your hardware

but you can try

turbo only need 1 step

lightning need 4 or 8 steps (i do not use 2 step variant)

sdxl lcm is smilar to sdxl 1.5 lcm, but produce bigger images

and works much slower

[karmic crown]

HTTPError: 401 Client Error: Unauthorized for url: https://huggingface.co/animeScreencap.safetensors/resolve/main/tokenizer/tokenizer_config.json

you are not allowed to access that file
401 is error that means that you are not authorized to access that

[primal glen]

so "lcm" and "lightning" are useful at speeding up the model, but SDXL is a slower model to begin with so I might be better off using a faster model like regular Stable Diffusion?

[karmic crown]

its like this

[primal glen]

I don't see any models that are LCM or lightning that aren't SDXL

[karmic crown]

SD 1.5 LCM

[thorny mortar]

I have a question

do I need to access that file ?

[karmic crown]

i have no idea, sorry 😦

[primal glen]

okay if I can't read I think that means it's time to take a break. Thanks for the suggestions, I'll be back with my results and probably more requests for help

[thorny mortar]

the animeScreencap.safetensors is a model file already in my computer

[karmic crown]

1.5 was trained on 512x512px, and it generates that by default
SDXL was trained on 1024x1024px, and it generates that by default

that is the reason why sdxl is slower than 1.5

lcm is the modified version that exists for both that requires fewer steps, so it is "faster", but only because that requires fewer steps

lightning is even faster but there is no lightning version of 1.5 models

[karmic crown]

https://discuss.huggingface.co/t/error-401-client-error-unauthorized-for-url/19714/38

looks like it requires authorization with token

do you have HF_TOKEN?

[thorny mortar]

no I dont

just confuse because if I use automatic 1111 never need that

[karmic crown]

are you sure that can work without it and offline, since it ask for protected resurece on HF?

[thorny mortar]

No i not sure about it, but in automatic 1111 I see it don't need connect to HF

[karmic crown]

try to open this link in browser
https://huggingface.co/animeScreencap.safetensors/resolve/main/tokenizer/tokenizer_config.json

you will see that it say invalid password

and that link is from your log

[thorny mortar]

yes, because path

animeScreencap.safetensors/...

[karmic crown]

better wait for someone who knows more about this, sorry

[thorny mortar]

this path is same as my checkpoint file

so that mean it not exist

[brave frigate]

can some help me with this?

im using AMD

[thorny mortar]

it say you out of memory

[brave frigate]

20 GB 7900 XT hellhound

idk how to fix it tho

[karmic crown]

have you used one of the amd tutorials from this channel to set it up?

[brave frigate]

not from this channel no.

[karmic crown]

in the pinned messages (top right) of this channel you have two amd tutorials

the one that use zluda is probably better for you

[brave frigate]

I'll try that out

[ornate elk]

Please use the AMD Zluda guide, you won't have any vram problems with that and its blazing fast

I have a 7900xtx, works very good with it

The guide is in the pinned messages of this channel

Feel free to ask if anything isnt clear

[tepid crypt]

is there any way to identify what model was used to train a lora?

[vocal burrow]

apart from reading the description page/readme/etc. no (to my knowledge)

[high nymph]

Hello,
it is possible to create 180 degree or 360 images?
Thanks

[iron yoke]

Hello, how do generate an image

[brave frigate]

I have done the ZLUDA setup now but now i get a new error..

venv "C:\Users\Chef\Desktop\SD-Zluda\stable-diffusion-webui-directml\venv\Scripts\Python.exe"
WARNING: ZLUDA works best with SD.Next. Please consider migrating to SD.Next.
Using ZLUDA in C:\ZLUDA
Python 3.10.11 (tags/v3.10.11:7d4cc5a, Apr 5 2023, 00:38:17) [MSC v.1929 64 bit (AMD64)]
Version: v1.9.0-6-gb7c5e603
Commit hash: b7c5e6039e3a60dfd0b4df342782fabdf6844a16
Skipping onnxruntime installation.
You are up to date with the most recent release.
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
C:\Users\Chef\Desktop\SD-Zluda\stable-diffusion-webui-directml\venv\lib\site-packages\pytorch_lightning\utilities\distributed.py:258: LightningDeprecationWarning: pytorch_lightning.utilities.distributed.rank_zero_only has been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it from pytorch_lightning.utilities instead.
rank_zero_deprecation(
Launching Web UI with arguments: --use-zluda --medvram-sdxl --update-check --skip-ort

rocBLAS error: Cannot read C:\Program Files\AMD\ROCm\5.7\bin/rocblas/library/TensileLibrary.dat: No such file or directory for GPU arch : gfx1036

rocBLAS error: Could not initialize Tensile host:
regex_error(error_backref): The expression contained an invalid back reference.
Press any key to continue . . .

i dont really know how i can make it read the correct things

[ornate elk]

you need to disable your integrated AMD gpu from your CPU, youll find it in the device manager under Graphicadapters.
if you have two entrys there disable the one that isnt the 7900xt

[fair oxide]

the endpoint isn't working? https://stablediffusionapi.com/api/v3/text2img

[ornate elk]

Nice 🙂

[balmy quest]

everytime i generate a img2img i get IndexError: list index out of range
i dont know what to do before it worked fine
i didnt use any extension i think

[sharp lichen]

what version of python should i use for comfyui?

[ornate elk]

3.10.11

[hard quiver]

hi. ive installed forge and its not opening up when I press anykey. any idea?

i also already have automatic 1111 installed.

maybe thats releated somehow?

[rocky adder]

Great! thanks for the response

[umbral aspen]

Hey all, total noob here.
Last night, I managed to get the stable diffusion webui (automatic1111/stable-diffusion-webui-1.0.0-pre) installed and working. I had to change/add a few requirements_versions and I am using the arguments --precision full --no-half to launch. I'm currently trying to get sd-webui-segment-anything to work, but running into issues after installing it. Is this the right place to ask for help?

Thanks!

[rocky adder]

What is Deforum Dischord? I just don't find it

[vocal burrow]

https://github.com/deforum-art/deforum-stable-diffusion?tab=readme-ov-file#customization

[ornate elk]

Yes, what's your GPU?

[umbral aspen]

Ancient and underpowered 😅
I think it's an rx480. Currently looking for a newer Nvidia based one.

[vocal burrow]

oof yeah you might be stuck with pytorch 1 using this

[thin glen]

What up guys, does anyone know where I can get help with animatediff in comfyui? I'm trying to find a motion module model that works with the SDXL refiner (if there is one)

[gaunt mural]

hey i get this masseage when i try to train my dreamboot model PermissionError: [Errno 13] Permission denied: 'trainingImages\attika'
any one now what i can do

[ornate elk]

Make sure you also add --opt-sub-quad-attention

And if you have 4gb vram then --lowvram

If 8gb then --medvram

[umbral aspen]

ok, I'll add those. The issue I'm having with installing/adding sd-webui-segment-anything looks like it might be path related.

Traceback (most recent call last):
  File "C:\StableDiffusion\stable-diffusion-webui-1.0.0-pre\modules\scripts.py", line 218, in load_scripts
    script_module = script_loading.load_module(scriptfile.path)
  File "C:\StableDiffusion\stable-diffusion-webui-1.0.0-pre\modules\script_loading.py", line 13, in load_module
    exec(compiled, module.__dict__)
  File "C:\StableDiffusion\stable-diffusion-webui-1.0.0-pre\extensions\sd-webui-segment-anything\scripts\api.py", line 10, in <module>
    from scripts.sam import sam_predict, dino_predict, update_mask, cnet_seg, categorical_mask
  File "C:\StableDiffusion\stable-diffusion-webui-1.0.0-pre\extensions\sd-webui-segment-anything\scripts\sam.py", line 21, in <module>
    from scripts.auto import clear_sem_sam_cache, register_auto_sam, semantic_segmentation, sem_sam_garbage_collect, image_layer_internal, categorical_mask_image
  File "C:\StableDiffusion\stable-diffusion-webui-1.0.0-pre\extensions\sd-webui-segment-anything\scripts\auto.py", line 12, in <module>
    from modules.paths import extensions_dir
ImportError: cannot import name 'extensions_dir' from 'modules.paths' (C:\StableDiffusion\stable-diffusion-webui-1.0.0-pre\modules\paths.py)```

surprisingly, it actually works fairly well and quickly for smaller images (512x512). I'm excited to see what a modern GPU will do

[wide badge]

Hey I think this is the right place to ask this question, not sure though.
I want to set up a webui and I was wondering if there are any webuis set up already loaded up with some of the new PAG tech.

like a github page or something y'all know of

If not, thanks for your time appreciate ya

[wind roost]

can someone help me with the API?

i cant define the sampelr

its annoying me

is this correct?

            "prompt": prompt,
            "steps": steps,
            "sampler": "DPM++ 3M SDE Karras",
            "override_settings": {
                "sd_model_checkpoint": model_checkpoint,
                
            }
        }

[frosty summit]

Is anyone else not able to see images when their UI when generated? I am able to see my generated images just fine in my output folder but I can't see them in my UI

[unique bluff]

Is sd supposed to take extra long to generate an image? It’s been over 15 min and still at 0% (6 min of that was just to get to the progress part)

It’s simply not moving forward

It stays stuck at 0%| | 0/25 [00:00<?, ?it/s] Total progress: 0it [00:00, ?it/s]

[ripe bronze]

Comfyui: I'm confused why this is showing as red. See in the bottom picture I already have that file in the custom nodes folder.

Same issue as stated above now with another red error message (and different workflow):

[scenic skiff]

No. It should not take that long. Something must be going wrong.

[sharp lichen]

Does anyone know how i could have controlnet only work on my subject? and have it ignore the rest of controlnet input image?

[hard shale]

Do you mean these repos?
webui https://github.com/v0xie/sd-webui-incantations
forge https://github.com/pamparamm/sd-perturbed-attention
These are extensions, but you can easily use them.

[tulip widget]

why I can't generate image in Stable Diffusion in Discord?

no bot room under my account

[vocal burrow]

cf #1047610792226340935

[serene robin]

Hi, my automatic111 on an AMD gpu is not working since they update something 3 months ago

how can i best unistall & delete it?

[ornate elk]

Hey what's your GPU?

[serene robin]

6700 xt 12gb

[vocal burrow]

Simply select its directory and delete it. But you might want to save your models, loras, embeddings, outputs, etc beforehand if you want to reinstall it. Check pinned message for that.

[ornate elk]

Okay, there is a new and faster webui for AMD cards like yours.
I made a install guide.
Its in the pinned messages of this channel

Zluda version

[serene robin]

good call, i'll save those!

[serene robin]

awesome! i'll try it

[ornate elk]

Sure, you can then easily use your existing models and Lora's from the older version.
Also if you have any questions feel free to ask. 🙂

[serene robin]

i saved VAEs too

thanks guys, i'll try that and i'll let you know once I install it ❤️

[ornate elk]

Also save upscalers if you had custom ones

They are in the models/ESRGAN folder

[serene robin]

i didn't have those

but do i need to save torch, because when i first installed this version i had a hard time making it work because of it

[ornate elk]

Nope, the zluda version needs a different version of torch. No need to save it.

[serene robin]

alright then

[placid robin]

https://github.com/Stability-AI/stability-sdk/issues/282

[solemn atlas]

Hii, in which folder does "poses" need to be

[unique bluff]

When I tried to make a “simple”image (just a prompt) it did just fine

When I use multiple prompts + negative prompts + controlenet + high res it doesn’t

How do I fix this?

[gaunt mural]

i get this message when i rty and train my Dreambooth model

PermissionError: [Errno 13] Permission denied: 'trainingImages\attika'

[open needle]

Hi, I just installed SD on my win11.
Here I have 2x files:
-> sd_xl_base_1.0.safetensors (I downloaded it here: https://huggingface.co/stabilityai/stable-diffusion-xl-base-1.0/tree/main)
-> v1-5-pruned-emaonly.safetensors (it came out-of-the-box)
C:\git\stable-diffusion-webui\models\Stable-diffusion

When I use SD, f.ex. for "a man" v1-5-pruned-emaonly.safetensors gives me a normal image, but sd_xl_base_1.0.safetensors gives me a weird looking image (like a Picasso art). Is that normal?

I tried these 2x models: https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Features
And this seems to work (I get normal images):
-> sd_xl_refiner_1.0_0.9vae.safetensors

So in short, these produce a normal image (not some Picasso art):
-> v1-5-pruned-emaonly.safetensors
-> sd_xl_refiner_1.0_0.9vae.safetensors
These produce Picasso art:
-> sd_xl_base_1.0_0.9vae.safetensors
-> sd_xl_base_1.0.safetensors

[ornate elk]

Hey, sdxl models need a base resolution of 1024x1024

1.5 models are trained on 512x512

[fierce hollow]

Please explain in simple terms the difference between CHECKPOINT TRAINED and CHECKPOINT MERGE.

[onyx solar]

Merged = multiple existing models smashed together
Trained = model created from scratch using images

[fierce hollow]

Thank you very much.

[tepid crypt]

someone knows what model is the one they mention here? im trying to use their Lora but cant find a model that fits good, hope someone can help
the Lora: https://civitai.com/models/18472

[onyx solar]

They're referring to NovelAI.

[tepid crypt]

is it possible to use it out of the site?

[onyx solar]

lol

That's a whole thing.

[ornate elk]

Use the AnythingV5 model

Closest you get to novelai

[tepid crypt]

got it, thx

[stone junco]

is there a way to use older versions of chat gpt?

[fierce hollow]

The Model 1.5 is trained on 512x512, I read it in this chat. Is there a way to increase these values?

[ornate elk]

you can try 768x768 or any other resolution near 512, but if you go to high you get duplicates and artefacts.
the best way is to upscale to a higher resolution with hires fix

[cosmic furnace]

Hmmm. Amd hip ray tracing here or almost here?)

[fierce hollow]

I'm a beginner trying to get the hang of it, and as you say, "artifacts" like giraffes with three legs, two arms, Siamese twins, and so on keep coming out.

[cosmic furnace]

I guess if only 3 games have signifficant bad ray tracing with 7900 xtx its something about those games optimisation? Like cyberpunk & hogwarts.

[ornate elk]

yea that can happen, also use a good negative prompt to not get artefacts

[ornate elk]

cyberpunk works good with raytraying on my 7900xtx, but i have also FSR enabled on Quality

[fierce hollow]

Where can I get these negative promts? Preferably universal, for people

[ornate elk]

here is a basic one:
blurry, disfigured, kitsch, ugly, oversaturated, grain, low-res, Deformed, blurry, bad anatomy, disfigured, poorly drawn face, mutation, mutated, extra limb, ugly, poorly drawn hands, missing limb, blurry, floating limbs, disconnected limbs, malformed hands, blur, out of focus, long neck, long body, ugly, disgusting, poorly drawn, childish, mutilated, , mangled, old, surreal, text, blurry, b&w, monochrome, conjoined twins, multiple heads, extra legs, extra arms, meme, deformed, elongated, twisted fingers, strabismus, blurred, watermark

[cosmic furnace]

btw have you seen THIS? https://youtu.be/_toA8lErAHg

Mind blowing y? I thought first 30 sec was a real life video 😄

trying to fool me

[fierce hollow]

Okay, I'll try it now, thanks. Are these sizes not suitable for 1.5?

aspect ratio: 4:3 - 1152x896
aspect ratio: 2:3 - 832x1216
aspect ratio: 9:16 - 768x1344
aspect ratio: 21:9 - 1536x640
aspect ratio: 1:1 - 1024x1024
aspect ratio: 3:4 - 896x1152
aspect ratio: 16:9 - 1344x768
aspect ratio: 5:4 - 1152x896
aspect ratio: 3:2 - 1216x832
aspect ratio: 4:5 - 896x1152
aspect ratio: 9:21 - 640x1536

[ornate elk]

yea its crazy

[ornate elk]

these are for sdxl only

[cosmic furnace]

Btw ive heard Sony wants their own Ray tracing. And with this AMD hip ray tracing i think amd trying to catch up? & it might happen. Cause open sourse like always hmgm

[fierce hollow]

Tell me, are there any sdxl models that support lora?

[ornate elk]

every sdxl model supports loras, but the loras also need to be trained on sdxl

[fierce hollow]

How to do it? If I conditionally connect the "Juggernaut XL" model, then Lora does not show any available plugins, as soon as I switch to the 1.5 Lora model, it is visible

[ornate elk]

yes, as in every console (PS or Xbox) is an AMD Chip, we will get better RT maybe
but i dont care about Raytraycing, its a gimmick which lowers my fps significant for a not that noticeable difference (depending on game)

[ornate elk]

that means the lora is made for 1.5
and wont work with sdxl

[cosmic furnace]

Did you see video that some enthusiast made an 7900xtx be more powerful than 4090 ? He somehow bypassed the power limits & crancked it up to have more 1.5x power to it running and making it about 15-20% better ;O

[fierce hollow]

So it already depends on the specific Lora model?

[cosmic furnace]

Gotta agree RT is kinda overrated =\

[ornate elk]

yes its possible xD

[ornate elk]

i dont even use it in Cyberpunk xD

max settings is enough on 2k monitor

looks awesome

[ornate elk]

yes

on civitai you can filter

for loras made for 1.5 or sdxl

[cosmic furnace]

PPl saying u need N-asa PC to run CyberPunk & photo real mod-presets but all i can see is 4090. I do know theres cpu that costs 10k $ but doesnt seems like the case

Did u tried this photoreal mod? ;o

[ornate elk]

nope, but i will mod my Cyperpunk in a few days/weeks, i have already a list with mods i need

[cosmic furnace]

Ray tracing affect overall picture of it much or nah btw? How dyu think?)

[ornate elk]

most is done by the Depth of field blur and the reshade

the game already has a good lighting so shadows should be okay

with raytracing they would be better, and of course some transparent textures or mirros

[cosmic furnace]

I wonder why amd has locked Power increase on 7x series =\ . Ppl was playing with it in 6x series too much? Lowers overall lifespan of a card? 🤔 Just a safety measure?

Weird decision if it makes AMD better than Nvidia in productivity

[ornate elk]

nope, AMD locks settings at values that are compatible and safe for every System
They also recommend to tweak as much as you like, best to do so is with the Adrenalin Software.

[cosmic furnace]

only +10% there 😦

and they say different brand have different highest&standart power limits . As if bios locked by manufacturer

[ornate elk]

yes, i have the 7900XTX Nitro+ Vapor-X by Saphire , it comes already overclocked

[cosmic furnace]

y they say those have highest power limits 🤔

[ornate elk]

@cosmic furnace checkout this video for adrenalin overclocking:
https://www.youtube.com/watch?v=xdCWdvW8qYk

[cosmic furnace]

really weird why diff brands make diff powa limits & they differ so much tho =\

[ornate elk]

they want to make a selling point ^^ if every third party card would have the same performance, you just would by the cheapest one

its the same for nvidia cards, third party cards are better than the ones shipped by nvidia

[cosmic furnace]

indeed 🙂

[edgy heath]

How to fix stable diffusion speed issues

mac m2 chip

[ornate elk]

whats the issue?

[storm lake]

Hi yall, im trying to get stable diffusion to work for the first time and it just kinda isnt
Im using this guide https://github.com/AUTOMATIC1111/stable-diffusion-webui?tab=readme-ov-file and the "automatic installation on widows" bit
I got everything installed i think, but now that im trying to run webui-user.bat i keep getting this: "RuntimeError: Torch is not able to use GPU; add--skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check" and "Press any key to continue" but it just closes when i do.
Idk very much when it comes to cmd and editing files and all that, what do i do to make it work?

[ornate elk]

Hey, what's your GPU?

[storm lake]

not sure actually where do i check
not a particularly good one im sure, im just trying it out on my laptop

[edgy heath]

2hours for 1 image

speed

[storm lake]

ok found it apparently its an Intel(R) UHD Graphics 620

[ornate elk]

Check in the device manager under graphics adatper

Or in the task manager under performance then on GPU

[storm lake]

^

[ornate elk]

Your sure?

[storm lake]

should be

[frosty summit]

Hey, has anyone else had the issue where you can’t see the image generated in the UI? I am able to access the image I generate in the output directory but I am not able to see it directly in the UI. Not sure if it’s a setting or something that I need to enable, but any help would be appreciated.

[ornate elk]

Then it would only work in CPU mode.
To get it working you have to edit the webui-user.bat and at the line commandline_args=
You need to add: --skip-torch-cuda-test --use-cpu all --opt-sub-quad-attention --no-half

[storm lake]

alr thx ill try

[ornate elk]

But recheck in taskmanager under performance tab. If you have GPU 0 or GPU 0 and 1

[storm lake]

just gpu 0

[ornate elk]

Okay yea then do that steps above

[storm lake]

it seems to be doing something rn at least, thx for the help

[ornate elk]

Did you used the Install script for macs from the pinned messages of this channel?
If not, you should try that

[hearty minnow]

**hey guys , in need of help i recive this kind of error **
And my AI dos't start

File "D:\AI\stable-diffusion-webui-forge\launch.py", line 51, in <module>
main()
File "D:\AI\stable-diffusion-webui-forge\launch.py", line 47, in main
start()
File "D:\AI\stable-diffusion-webui-forge\modules\launch_utils.py", line 541, in start
import webui
File "D:\AI\stable-diffusion-webui-forge\webui.py", line 19, in <module>
initialize.imports()
File "D:\AI\stable-diffusion-webui-forge\modules\initialize.py", line 35, in imports
import gradio # noqa: F401
File "D:\AI\stable-diffusion-webui-forge\venv\lib\site-packages\gradio_init_.py", line 3, in <module>
import gradio.components as components
File "D:\AI\stable-diffusion-webui-forge\venv\lib\site-packages\gradio\components_init_.py", line 1, in <module>
from gradio.components.annotated_image import AnnotatedImage
File "D:\AI\stable-diffusion-webui-forge\venv\lib\site-packages\gradio\components\annotated_image.py", line 13, in <module>
from gradio.components.base import IOComponent, Keywords
File "D:\AI\stable-diffusion-webui-forge\venv\lib\site-packages\gradio\components\base.py", line 20, in <module>
from fastapi import UploadFile
File "D:\AI\stable-diffusion-webui-forge\venv\lib\site-packages\fastapi_init.py", line 7, in <module>
from .applications import FastAPI as FastAPI
File "D:\AI\stable-diffusion-webui-forge\venv\lib\site-packages\fastapi\applications.py", line 15, in <module>
from fastapi import routing
File "D:\AI\stable-diffusion-webui-forge\venv\lib\site-packages\fastapi\routing.py", line 22, in <module>
from fastapi import params
File "D:\AI\stable-diffusion-webui-forge\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' (D:\AI\stable-diffusion-webui-forge\venv\lib\site-packages\pydantic\fields.py)
Press any key to continue . . .

i reinstalled ai , with removing python and git completley , but still recive same msg

[tough delta]

Hey. I'm getting this error with Kohya.

local variable 'epoch' referenced before assignment

https://pastebin.com/Gxx62Dtv

I'm trying to learn to make some lora all day but I only getting errors even tho I followed youtube videos about installing etc. Currently I'm trying "Kohya-v22.6.2" as suggested because newer versions are a bit of mess and gives even more errors.

ahhhhh this is crazy. I see some progress now. I fixed bit and something, but next time I tried running via dreambooth instead lora tab and gave me error.

[edgy heath]

https://github.com/viking1304/a1111-setup
this im assuming. no i did not. in fact, i have multiple stable diffusion folders both downloaded differently. i was drunk and new to the terminal

do i uninstall stable diffusion and reinstall? if so - how to uninstall?

[ornate elk]

uninstall is easy, just delete the stable-diffusion-webui folder
but save your outputs, models and loras if you had them

[edgy heath]

Cheers

[ornate elk]

Then run the script, it will install SD with the best performance settings

[hearty minnow]

could you look at my screenshot please ?

[ornate elk]

Yea thats a common issue right now.
You can fix it by following this:
https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/15551#issuecomment-2061997952
Also here is an workaround too:
https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/15564

[hearty minnow]

will try it right now

[hearty minnow]

fixed ❤️

[cosmic furnace]

Not sure you can do this. But i guess if you change bios firmware of 1 gpu that has lower limits from other that have higher power limits you can rise max powa limits? 🤔

Ive heard there was an bios firmware of nvidia 4090 or 3090 leaked & ppl can now install it on any lower. Gaining some bettter performance and making system think that its 90 series 😄

[drowsy creek]

well RIP my gpu doesn't support pytorch

(RX 5600 XT)

so is locally running SD impossible now on arch?

I was able to run it on CPU I think on windows

how would I do that on arch?

[ornate elk]

Hey a 5600xt will work on windows better now. In the pinned messages of this channel you find the SD webui zluda guide.

No need to run it on CPU

[drowsy creek]

I was able to run it on CPU on windows *in the past

I fully run linux rn

and the RX 5600 XT was not on that page I highlighted

[ornate elk]

yea than you would have to build it for yourself, but you could also try to get it run using Zluda but on linux

idk how but it should be possible

[drowsy creek]

is Zlunda a seperate webui than amd1111?

or is it a torch compatability layer for amd1111?

[ornate elk]

zluda is a compatibility layer for cuda to work with AMD gpus.
it works as backend for auto1111 like directml or rocm

your card is supported by zluda

@drowsy creek
Here are the linux and windows files for zluda:
https://github.com/lshqqytiger/ZLUDA/releases/tag/rel.9e97c717c3fef536d3116f39a15d95626c1dfe39
Here are the comunity made Rocm libs that support older AMD gpus like yours.
https://github.com/brknsoul/ROCmLibs/

you need Rocm installed and the latest AMD drivers

[drowsy creek]

https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Install-and-Run-on-AMD-GPUs#install-on-amd-and-arch-linux

where would that fit into this tutorial

what step?

[thin glen]

Is there a motion module that's compatible with the SDXL refiner? trying to create a vid2vid workflow with animatediff that works with SDXL.

or do I just have to run the whole thing using the base model

[ornate elk]

hmm after setting up the venv i guess
then you have to install pytorch for cuda inside this venv
torch 2.2.2 with cuda 11.8

[wraith pewter]

OutOfMemoryError: CUDA out of memory. Tried to allocate 18.86 GiB. GPU 0 has a total capacity of 23.98 GiB of which 0 bytes is free. Of the allocated memory 7.31 GiB is allocated by PyTorch, and 18.04 GiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables)

Why am I getting this

the image was 1 second before complete...

[wraith pewter]

oh it seems i gotta restart after a large image

[ornate elk]

Whats your GPU?

You may need the tiled VAE extension or the sdxl fp 16 vae

[strange fiber]

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. i dont know how to fix it someone can hellp^?

[ornate elk]

Hey, what's your GPU?

[strange fiber]

AMDryzen 53500x6core

[ornate elk]

Thats your CPU

[strange fiber]

NVIDIA 1660 super

[ornate elk]

ah okay, then you need to edit your webui-user.bat. at the line COMMANDLINE_ARGS=
you need to add: --xformers --medvram --no-half
then save and relaunch it

[strange fiber]

TY

[main owl]

OutOfMemoryError: CUDA out of memory. Tried to allocate 5.37 GiB. GPU 0 has a total capacty of 15.99 GiB of which 0 bytes is free. Of the allocated memory 21.94 GiB is allocated by PyTorch, and 3.34 GiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF, Stable Video Diffusion is there any guide or tip for it?

[ornate elk]

Hey what's your GPU?

[main owl]

RTX 4080 16GB

[ornate elk]

Okay and what's on your webui-user.bat ?

[main owl]

downloaded from pinokio

[ornate elk]

Ah okay.
Then go into the stable-diffusion-webui folder.
Edit the webui-user.bat and there add the line commandline_args=
And there add: --xformers

[main owl]

there is no webui-user.bat

😄

[ornate elk]

Ohhh

[main owl]

i said its downloaded from pinokio app

[ornate elk]

Then I can't help much. But would recommend using the normal installation instead of Pinokio

[main owl]

well i will try later

[ornate elk]

In the pinned messages of this channel you'll find my Nvidia install guide

[tight dirge]

I am having the same issue, did you find a solution to this ?

[sinful needle]

I'm trying to set up Animate Diff. Where do I git clone the folder? In the stable diffusion folder? https://github.com/guoyww/AnimateDiff/blob/main/__assets__/docs/animatediff.md

[narrow copper]

Does comfy ui have anything to show prompt progress? There's the queue view but it doesn't show if anything's happening, just that something's running.

[onyx solar]

I recommend rgthree's node pack for that (among other nice things).
https://github.com/rgthree/rgthree-comfy

[narrow copper]

Lovely. Thanks. 😗

[sinful needle]

😦

[ripe bronze]

On comfui sometimes the Prompt Box auto-suggests loras. But sometimes it suddenly stop trying to autopopulate them. How do I keep it to always autopopulate loras that I have in my folder?

[drowsy creek]

how would I install the exact versions on arch? [EDIT: not gonna delete this message since I don't wanna ghost ping, but nvm]

[dapper basin]

cannot launch stable Diffusion

[vocal burrow]

1/ console logs ?
2/ what client/UI are you using ?
3/ you forgot to put the port at the end http://127.0.0.1:7860/

[north tundra]

I re-installed the windows and webui but keep the old models, extensions in another disc. After re-installing the webui with no problem, I just copy the every single old models and extensions into the new folder. Now I am having an issue with selecting a model. Anyone have an idea to fix it? Stable diffusion model failed to load changing setting sd_model_checkpoint to ohwxRVB2.ckpt: AttributeError Traceback (most recent call last): File "D:\Stable Diffusion\stable-diffusion-webui\modules\options.py", line 165, in set option.onchange() File "D:\Stable Diffusion\stable-diffusion-webui\modules\call_queue.py", line 13, in f res = func(*args, **kwargs) File "D:\Stable Diffusion\stable-diffusion-webui\modules\initialize_util.py", line 181, in <lambda> shared.opts.onchange("sd_model_checkpoint", wrap_queued_call(lambda: sd_models.reload_model_weights()), call=False) File "D:\Stable Diffusion\stable-diffusion-webui\modules\sd_models.py", line 860, in reload_model_weights sd_model = reuse_model_from_already_loaded(sd_model, checkpoint_info, timer) File "D:\Stable Diffusion\stable-diffusion-webui\modules\sd_models.py", line 793, in reuse_model_from_already_loaded send_model_to_cpu(sd_model) File "D:\Stable Diffusion\stable-diffusion-webui\modules\sd_models.py", line 662, in send_model_to_cpu if m.lowvram: AttributeError: 'NoneType' object has no attribute 'lowvram'

I am using A1111 btw

[dapper basin]

1. Don't know what those are
2. Default one probably. I just installed it the other day using this server's pinned messege
3. what's a port?

oh if the port means that 7860 part
ya it still doesn't work with that

[vocal burrow]

console logs, it's the text written in white in the black window thingy opens up when you launch webui-user.bat

[dapper basin]

ok Not gonna lie
I googled how do i open sd after installation and it just said type in that address

it's working fine now
thanks for the help 😭

[dapper basin]

so i was testing out how gpu does sd use
when trying to create a max res image (2048x2048)
both dedicated and shared memory maxed out and I closed the generation at about 50%
does this impact my gpu/laptop in a bad way if I do this
obv. I am not gonna make full res images anyway but I just wanna how it affects my pc

[north tundra]

any idea?

[fair oxide]

is there a way to make stable diffusion slow down/use less resources/whatever for the purpose of not heating my laptop as much

[tiny garden]

adding to the sorting of LoRAs:
can I apply the same folder structure to the checkpoints (= models) as well?

[heavy perch]

mov2mov tab is not coming up on my webui, what should I do?

[solemn moon]

You good
Except cache sometimes

[solemn moon]

Restart it

If not check settings tab
Maybe /extensions or the like 👍

[unique bluff]

Is there any way tot fix the issue that happens when I try to make images and it doesn’t goes past 0%?

[ornate elk]

What's your GPU?

[heavy perch]

here?

[ornate elk]

Here is the fix for mov2mov:
https://github.com/Scholar01/sd-webui-mov2mov/issues/146#issuecomment-2054157728

[heavy perch]

how do I set it up tho
super new to this

[ornate elk]

You already have it installed. But you need to do the steps in the fix

[heavy perch]

which one you mean

[unique bluff]

It’s the intel iris pro graphics 6200 1536 MB, but I think I’ve been running on cpu mode only

[ornate elk]

The top one

With the zip

[ornate elk]

Your the Mac user right?

What does the cmd show when you launch the webui?

[heavy perch]

ok where do I put this?

[ornate elk]

I think it goes into the extensions mov2mov foldee

But I'm not at the PC to verify that

[heavy perch]

still didn't work

[cosmic furnace]

how smone has +15% power limit increase not +10% ? o.0 On adrenaline . Is it dependable silent-gaming mode on gpus or whuht?

[edgy heath]

Hey C! I reinstalled stable difusion and i cut it from 2 hours down to 3-5 minutes. Do you have any recommendations for further speed optimization for mac?

[dense tapir]

Hi! I couldn't replicate the Checkpoint merger step.

When I renamed mm_sdxl_v10_beta.ckpt to mm_sdxl_v10_beta.safetensor, I put it into the checkpoint folder and it wouldn't appear

[fair oxide]

@ornate elk still want help with this. i know youve given me good advice previously

[ornate elk]

Great to hear, The install script should already set the best performance args for Mac.
But maybe @karmic crown knows some tricks.

[ornate elk]

You need to put it back into animatediff folder after the merge

[ornate elk]

Hey, what was your GPU again?

[dense tapir]

I couldn't merge it that's the problem

I think I might be too dumb dumb to convert it to safetensor?

[ornate elk]

Maybe xD
If you put the .ckpt into models/stable-diffusion
You have to reload the checkpoints in txt2img.
Then it should be at the bottom in the dropdown

In the merger tab

[fair oxide]

rtx 3070 laptop

[drowsy summit]

im having the torch is not able to use gpu error, already installed the most recent version of pip

[dense tapir]

Good instructions thank you. So when I press Merge, the motion module file is still a CKPT in the models > Stable-diffusion folder

It produced a .json file right below it, hmm...

[ornate elk]

Read my instructions again from the last time

I said what to select

[ornate elk]

What's your GPU?

[ornate elk]

So you have --xformers --medvram-sdxl --no-half-vae in the webui-user.bat?

[drowsy summit]

im using amd

[ornate elk]

Which GPU model?

[drowsy summit]

not nvidia

AMD Radeon Pro 5300M

its updated im pretty sure

[fair oxide]

yes just double checked that

does that mean its set for moderate/safe usage?

its pretty warm right now

[dense tapir]

I believe I followed it step by step and it didn't give me a safetensor file

[drowsy summit]

should i just skip the test

for torch

[ornate elk]

No, for AMD you have to follow my AMD install guide in the pinned messages of this channel.
And not the Zluda guide.
This one you need:
#🤝｜tech-support message

[karmic crown]

What is your exact Mac model?

[drowsy summit]

will try that thank u

[ornate elk]

If you got a json file you didn't

[edgy heath]

apple pro m2 15gb

[ornate elk]

@dense tapir I can send you a screenshot later

[karmic crown]

post your command line params from webui-user.sh (or just paste the whole file)

[edgy heath]

.

[karmic crown]

just drag & drop file here

or use 3 backticks before and after

[dense tapir]

Oh sorry no need, I figured it out, thank you for your help, rn I'm testing out AnimateDiff with an XL model!

[grim moon]

Hi, prob really stupid question. but can stable diffusion expand an image with ai? Does it use models to do that?

[dense tapir]

For some reason it really didn't want to convert the model to safetensors and all I had to do was restart the cmd

[karmic crown]

below this line

#export COMMANDLINE_ARGS=""

add this

export COMMANDLINE_ARGS="--skip-torch-cuda-test --opt-sub-quad-attention --upcast-sampling --no-half-vae --use-cpu interrogate"

it should help with speed

[edgy heath]

gracias you sexy thang

[quick forge]

so a while back I tried running ForgeUI and long story short something went vastly wrong, and completely messed up my stable diffusion install.
I've tried reinstalling A1111 but it just errors out even when a completely fresh install.
I have no idea what happened so I don't even know what might need to be reinstalled to get it working again.
,,, error generated when compiling for gfx1030.
terminate called after throwing an instance of 'miopen::Exception'
what(): /long_pathname_so_that_rpms_can_package_the_debug_info/data/driver/MLOpen/src/hipoc/hipoc_program.cpp:304: Code object build failed. Source: naive_conv.cpp
webui.sh: line 297: 6167 Aborted (core dumped) "${python_cmd}" -u "${LAUNCH_SCRIPT}" "$@"
,,,

frig i did that wrong.

[karmic crown]

when you start webui, what it said at top about python version?

this part

################################################################
Launching launch.py...
################################################################
Python 3.10.14 (main, Mar 19 2024, 21:46:16) [Clang 15.0.0 (clang-1500.3.9.4)]
Version: f0.0.17v1.8.0rc-latest-276-g29be1da7

[quick forge]

How do you do that text box?

[karmic crown]

3 backticks before and after

[quick forge]

Launching launch.py...
################################################################
glibc version is 2.35
Check TCMalloc: libtcmalloc_minimal.so.4
libtcmalloc_minimal.so.4 is linked with libc.so,execute LD_PRELOAD=/lib/x86_64-linux-gnu/libtcmalloc_minimal.so.4
Python 3.10.12 (main, Nov 20 2023, 15:14:05) [GCC 11.4.0]
Version: v1.9.0
Commit hash: adadb4e3c7382bf3e4f7519126cd6c70f4f8557b

[karmic crown]

1.9.0 is normal 1111, i though you write its forge
i am confused now

[quick forge]

Forge is what I tried to run that fked up everything

im just trying to get A111 running again now

[grim moon]

Technicly speaking, can amd cards do img to img? My has some problems, it generated this error File "C:\Users\Dominik\sd\webui\stable-diffusion-webui-directml\modules\devices.py", line 152, in test_for_nans raise NansException(message) modules.devices.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

[quick forge]

I already completely reinstalled linux from scratch

[karmic crown]

fastest solution would be to rename the folder with sd (DO NOT delete it, just rename it)
install JUST a1111 (you have proper python and git)
and then move models and loras from old to new install

[quick forge]

Thats exactly what I've done

i didn't even move anything

i literally just completely recloned the repo and ran webui.sh

well other than pointing to my venv folder which may be the problem, but if I don't it won't automatically create it like it has before.

[karmic crown]

do not use old venv

let it create a new one

if venv does not exist a1111 will create it, so do not worry about it

but it can be a mess if it is corupted

thats why my a1111-setup script for mac always delete venv

[quick forge]

Create and activate python venv
################################################################

################################################################
ERROR: Cannot activate python venv, aborting...
################################################################

thats what I get when I let it try to create it on it's own.

[fair oxide]

i wonder if you forgot to tick mark Add path to python when re-installing python

[quick forge]

hmmm.... it's possible?

[fair oxide]

a common mistake ppl make

[quick forge]

how can i fix thst if i didn't?

[ornate elk]

A perfect 🙂

[ornate elk]

What's your GPU?

[grim moon]

RX6900XT

Thanks for replying

[fair oxide]

by reinstalling python and you should see the add path option at the first dialog box at bottom

[grim moon]

My args are : set COMMANDLINE_ARGS= --medvram --opt-sub-quad-attention --opt-split-attention-v1 --no-half-vae --upcast-sampling

[ornate elk]

Yea it should take the stress out of generation.
But laptops GPUs always get a lot hotter than PC GPUs.
Its okay.
But if its to hot you should buy a coolpad and make sure the fan intakes of the laptop (mostly at the bottom or behind) are not covered or blocked by anything

[quick forge]

I'm on linux, there was no installer like windows

[fair oxide]

oh snap, im not familiar with linux environment to install python, but its possible add path is missing

unless linux doesnt require add path, but your error message kinda hinted me that about path

[ornate elk]

Ah okay, so your using the old AMD webui Version which is pretty limited in vram and performance.

You need to follow the new Zluda AMD guide.
You'll find it in the pinned messages of this channel.
With that you can do mostly everything. Not limiting on vram and its much faster.

[grim moon]

should i delete my sd folder before install?

[ornate elk]

First save any models and Lora's and your outputs if you have used it.
If it was a fresh install you can delete it

[grim moon]

okay thanks

[karmic crown]

what you get with

python --version
python3 --version

[grim moon]

May I ask what DirectML webui is?

[karmic crown]

@quick forge
if i am right and python --version does not return 3.10
find this line in webui-user.sh

# python3 executable
#python_cmd="python3"

and uncomment the second line to became

# python3 executable
python_cmd="python3"

that it should create venv without any problem

[ornate elk]

The automatic1111 webui for AMD

[grim moon]

what is that?

[ornate elk]

The one you installed

before

[grim moon]

But because i deleted it, it doesnt matter to me, right? There arent any hidden files left?

[ornate elk]

Yes

[grim moon]

Thank you very much

[ornate elk]

If you want to start clean you can open up a cmd and type
Pip cache purge

[grim moon]

thanks, did it. Maybe add it to the pin, so no one will ask

[quick forge]

python --version
Python 3.10.12

[karmic crown]

that looks fine

it should create venv

[quick forge]

now i get this suddenly

Create and activate python venv
################################################################
The virtual environment was not created successfully because ensurepip is not
available.  On Debian/Ubuntu systems, you need to install the python3-venv
package using the following command.

    apt install python3.10-venv

You may need to use sudo with that command.  After installing the python3-venv
package, recreate your virtual environment.

Failing command: /media/killavolt/STABLE DIFFUSION/Front Ends/Automatic1111/venv/bin/python3

[karmic crown]

do sodo apt install python3.10-venv

hmm

[quick forge]

before I do that...... I didn't have to do that before. As you said, A1111 has always just automatically created venv, I don't know why it isn't now.

[karmic crown]

those are two diffrent things

you need python3.10-venv to able to create venv folder

i was talking about venv folder

venv folder is virtual enviroment with a copy of pyton files and packages installed just for that enviroment

but to be able to created you need a python3.10-venv (app)

[quick forge]

yeah im just saying that isn't a command I had to run before, when I installed this in the past.
The tutorial I followed though now seems to be gone from civitai so I can't lookup if I missed a command there.

I ran the command and am still getting the error

ERROR: Cannot activate python venv, aborting...

oh wait

had to delete the venv folder that was half created.

[karmic crown]

yes

[quick forge]

seems to be doing something now so lets see if it will finally run

I still have no idea what happened, I was doing a generation with forgeui, then suddenly my entire system locked up and apparently a bunch of stuff got corrupted somehow.

[karmic crown]

i only have virtual machines with linux, so i cant help much

i do not have any linux with GPU

[quick forge]

it's friggin working again 🙂

[karmic crown]

i installed a1111 on my vps, just to see would it work, but it is painfully slow with CPU only, even it has 10 CPUs and 60GB RAM 🙂

[karmic crown]

🎉

[wraith pewter]

@ornate elk this the one I need?

[quick forge]

let's see if it was just the venv folder that got messed up then and if my previoius install of a1111 is still good.

[karmic crown]

looks good

[grim moon]

Hey, is there any way you could help me on private dms? Im unsure about the path step, and i dont want to destroy my pc.

[unique bluff]

One sec

[karmic crown]

just move models

[quick forge]

my install is a bit customized

[ornate elk]

Yes

[karmic crown]

replace the file in scripts folder inside mov2mov extension

[unique bluff]

What part matters the most about the code when the ui launches?

[ornate elk]

Sure np

[wraith pewter]

do i need to do anything with it or can i just start creating larger images now

[karmic crown]

you need to use command line args optimized for your system

[ornate elk]

You only need to activate the Tiled VAE Option, thats all

[wraith pewter]

lol this is rlly pushing my pc to the max

[unique bluff]

It says this

################################################################
Install script for stable-diffusion + Web UI
Tested on Debian 11 (Bullseye), Fedora 34+ and openSUSE Leap 15.4 or newer.
################################################################

################################################################
Running on sweet.cake user
################################################################

################################################################
Repo already cloned, using it as install directory
################################################################

################################################################
Create and activate python venv
################################################################

################################################################
Accelerating launch.py...
################################################################
The following values were not passed to `accelerate launch` and had defaults used instead:
    `--num_processes` was set to a value of `0`
    `--num_machines` was set to a value of `1`
    `--mixed_precision` was set to a value of `'no'`
    `--dynamo_backend` was set to a value of `'no'`
To avoid this warning pass in values for each of the problematic parameters or run `accelerate config`.
Python 3.10.14 (main, Mar 20 2024, 03:51:51) [Clang 14.0.0 (clang-1400.0.29.202)]
Version: v1.9.0
Commit hash: adadb4e3c7382bf3e4f7519126cd6c70f4f8557b
ControlNet init warning: Unable to install insightface automatically. Please try run `pip install insightface` manually.
Launching Web UI with arguments: --skip-torch-cuda-test --no-half --use-cpu all
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
Warning: caught exception 'Torch not compiled with CUDA enabled', memory monitor disabled```

Then

==============================================================================
You are running torch 2.1.0.
The program is tested to work with torch 2.1.2.
To reinstall the desired version, run with commandline flag --reinstall-torch.
Beware that this will cause a lot of large files to be downloaded, as well as
there are reports of issues with training tab on the latest version.

Use --skip-version-check commandline argument to disable this check.
==============================================================================
ControlNet preprocessor location: /Users/sweet.cake/stable-diffusion-webui/extensions/sd-webui-controlnet/annotator/downloads
2024-04-21 15:24:03,102 - ControlNet - INFO - ControlNet v1.1.443
2024-04-21 15:24:03,441 - ControlNet - INFO - ControlNet v1.1.443
Loading weights [876b4c7ba5] from /Users/sweet.cake/stable-diffusion-webui/models/Stable-diffusion/cetusMix_Whalefall2.safetensors
2024-04-21 15:24:05,751 - ControlNet - INFO - ControlNet UI callback registered.
Running on local URL:  http://127.0.0.1:7860

To create a public link, set `share=True` in `launch()`.
Startup time: 68.6s (prepare environment: 4.0s, import torch: 12.5s, import gradio: 9.5s, setup paths: 10.9s, initialize shared: 2.4s, other imports: 8.8s, list SD models: 0.5s, load scripts: 3.9s, initialize extra networks: 0.5s, create ui: 3.5s, gradio launch: 12.4s, add APIs: 0.5s).
Creating model from config: /Users/sweet.cake/stable-diffusion-webui/configs/v1-inference.yaml```

[karmic crown]

ok

[unique bluff]

Keep in mind I did change the webui-user file because I wasn’t able to generate any images at all because of the lowvram error

[karmic crown]

first do this

venv/bin/pip install torch==2.1.2 torchvision==0.16.2

[unique bluff]

Now any more detailed images it doesn’t goes past 0% (only single prompts images load)

[unique bluff]

On the terminal?

[karmic crown]

yes, from stabel diffusion folder

then open webui-user.sh and change command line params to be like this

export COMMANDLINE_ARGS="--skip-torch-cuda-test --opt-sub-quad-attention --upcast-sampling --no-half-vae --use-cpu interrogate"

[unique bluff]

That when my computer does not restart on its own while launching sd

[unique bluff]

Humm, will try that

[heavy perch]

which

[quick forge]

I moved models over and it fked everything up again :/

[ornate elk]

in extensions/mov2mov

[heavy perch]

did that didn't work

[karmic crown]

pytorch is dropping support for intel macs, so i will update my script soon, it will do different things for silicon and intel macs

[unique bluff]

Nooooo

Tbh makes sense, my computer is an intel mac and is 10 years old lol

[karmic crown]

you will be able to use a1111, but you will not be able to use newer pytorch

[quick forge]

how do i install libstdc++-12-dev

nvm got it

[karmic crown]

sudo apt install libstdc++-12-dev i guess

@unique bluff did you managed to install newer version of torch and to run webui with new params?

[unique bluff]

I’m doing that now

My computer restarted on its own the last time I opened webui to copy the code

[karmic crown]

take your time

i need to go, but i will check later

[unique bluff]

Now it’s turning on again

[unique bluff]

It’s downloading now

[quick forge]

anyone know what this error is ?

*** Error calling: /media/killavolt/STABLE DIFFUSION/Front Ends/Automatic1111/modules/processing_scripts/sampler.py/ui
    Traceback (most recent call last):
      File "/media/killavolt/STABLE DIFFUSION/Front Ends/Automatic1111/modules/scripts.py", line 547, in wrap_call
        return func(*args, **kwargs)
      File "/media/killavolt/STABLE DIFFUSION/Front Ends/Automatic1111/modules/processing_scripts/sampler.py", line 20, in ui
        sampler_names = [x.name for x in sd_samplers.visible_samplers()]
    AttributeError: module 'modules.sd_samplers' has no attribute 'visible_samplers'

---
2024-04-21 15:09:06,661 - ControlNet - INFO - ControlNet UI callback registered.
*** Error calling: /media/killavolt/STABLE DIFFUSION/Front Ends/Automatic1111/modules/processing_scripts/sampler.py/ui
    Traceback (most recent call last):
      File "/media/killavolt/STABLE DIFFUSION/Front Ends/Automatic1111/modules/scripts.py", line 547, in wrap_call
        return func(*args, **kwargs)
      File "/media/killavolt/STABLE DIFFUSION/Front Ends/Automatic1111/modules/processing_scripts/sampler.py", line 20, in ui
        sampler_names = [x.name for x in sd_samplers.visible_samplers()]
    AttributeError: module 'modules.sd_samplers' has no attribute 'visible_samplers'

[karmic crown]

try to avoid spaces
rename
STABLE DIFFUSION
to something
STABLE_DIFFUSION

and only restore models, do not restore extensions (install them again)

[unique bluff]

I did that but now the lowvram error returned

And I can’t generate images at all

[karmic crown]

send your webui-user.sh here

let me check it

[unique bluff]

#########################################################
# Uncomment and change the variables below to your need:#
#########################################################

# Install directory without trailing slash
install_dir="/home/$(whoami)"

# Name of the subdirectory
clone_dir="stable-diffusion-webui"

# Commandline arguments for webui.py, for example: export COMMANDLINE_ARGS="--medvram --opt-split-attention"
export COMMANDLINE_ARGS="--skip-torch-cuda-test --opt-sub-quad-attention --upcast-sampling --no-half-vae --use-cpu interrogate"

# python3 executable
python_cmd="python3"

# git executable
GIT="git"

# python3 venv without trailing slash (defaults to ${install_dir}/${clone_dir}/venv)
venv_dir="venv"

# script to launch to start the app
LAUNCH_SCRIPT="launch.py"```

That’s with code you mentioned

[karmic crown]

that is a wrong file

webui-user.sh
not webui.sh

[unique bluff]

It’s the webui-user.sh file

Aaah

Oh

[karmic crown]

do
git restore webui.sh
to restore original webui.sh

[unique bluff]

It’s too many characters for my non nitro account, one sec

#################################################
# Please do not make any changes to this file,  #
# change the variables in webui-user.sh instead #
#################################################

SCRIPT_DIR=$( cd -- "$( dirname -- "${BASH_SOURCE[0]}" )" &> /dev/null && pwd )


# If run from macOS, load defaults from webui-macos-env.sh
if [[ "$OSTYPE" == "darwin"* ]]; then
    if [[ -f "$SCRIPT_DIR"/webui-macos-env.sh ]]
        then
        source "$SCRIPT_DIR"/webui-macos-env.sh
    fi
fi

# Read variables from webui-user.sh
# shellcheck source=/dev/null
if [[ -f "$SCRIPT_DIR"/webui-user.sh ]]
then
    source "$SCRIPT_DIR"/webui-user.sh
fi

# If $venv_dir is "-", then disable venv support
use_venv=1
if [[ $venv_dir == "-" ]]; then
  use_venv=0
fi

# Set defaults
# Install directory without trailing slash
if [[ -z "${install_dir}" ]]
then
    install_dir="$SCRIPT_DIR"
fi

# Name of the subdirectory (defaults to stable-diffusion-webui)
if [[ -z "${clone_dir}" ]]
then
    clone_dir="stable-diffusion-webui"
fi

# python3 executable
if [[ -z "${python_cmd}" ]]
then
    python_cmd="python3"
fi

# git executable
if [[ -z "${GIT}" ]]
then
    export GIT="git"
else
    export GIT_PYTHON_GIT_EXECUTABLE="${GIT}"
fi

# python3 venv without trailing slash (defaults to ${install_dir}/${clone_dir}/venv)
if [[ -z "${venv_dir}" ]] && [[ $use_venv -eq 1 ]]
then
    venv_dir="venv"
fi

if [[ -z "${LAUNCH_SCRIPT}" ]]
then
    LAUNCH_SCRIPT="launch.py"
fi

# this script cannot be run as root by default
can_run_as_root=0

# read any command line flags to the webui.sh script
while getopts "f" flag > /dev/null 2>&1
do
    case ${flag} in
        f) can_run_as_root=1;;
        *) break;;
    esac
done ```

[karmic crown]

your webui-user should look similar to this

[unique bluff]

export ERROR_REPORTING=FALSE

# Do not reinstall existing pip packages on Debian/Ubuntu
export PIP_IGNORE_INSTALLED=0

# Pretty print
delimiter="################################################################"

printf "\n%s\n" "${delimiter}"
printf "\e[1m\e[32mInstall script for stable-diffusion + Web UI\n"
printf "\e[1m\e[34mTested on Debian 11 (Bullseye), Fedora 34+ and openSUSE Leap 15.4 or newer.\e[0m"
printf "\n%s\n" "${delimiter}"

# Do not run as root
if [[ $(id -u) -eq 0 && can_run_as_root -eq 0 ]]
then
    printf "\n%s\n" "${delimiter}"
    printf "\e[1m\e[31mERROR: This script must not be launched as root, aborting...\e[0m"
    printf "\n%s\n" "${delimiter}"
    exit 1
else
    printf "\n%s\n" "${delimiter}"
    printf "Running on \e[1m\e[32m%s\e[0m user" "$(whoami)"
    printf "\n%s\n" "${delimiter}"
fi

if [[ $(getconf LONG_BIT) = 32 ]]
then
    printf "\n%s\n" "${delimiter}"
    printf "\e[1m\e[31mERROR: Unsupported Running on a 32bit OS\e[0m"
    printf "\n%s\n" "${delimiter}"
    exit 1
fi

if [[ -d .git ]]
then
    printf "\n%s\n" "${delimiter}"
    printf "Repo already cloned, using it as install directory"
    printf "\n%s\n" "${delimiter}"
    install_dir="${PWD}/../"
    clone_dir="${PWD##*/}"
fi```

[karmic crown]

let me send you both files

just replace them

[unique bluff]

Okay

[unique bluff]

I had to make some adjustments

My og didn’t have launch’s args

[unique bluff]

Nor vairables

[karmic crown]

try with those two

[unique bluff]

I asked ChatGPT and it gave me the code with the extra things needed lol

[karmic crown]

if that does not work, let me know what you get to send you new webui-user.sh

[unique bluff]

Now I don’t have permission to open the webui

idk why

[karmic crown]

chmod 755 webui.sh

did that help?

[unique bluff]

Yes

It’s opening

Wow, much faster than usual as well

[karmic crown]

let me know if you still get errors, so we can try other params

[unique bluff]

Trying to generate an image now

Without controlnet but with multiple prompts

[karmic crown]

start with basic generation

when you see that works, you can try more complicated things

you can try to use tiled vae, that extension might help with lowram machines

[unique bluff]

Oh no

Now two errors

[karmic crown]

show me

[unique bluff]

AttributeError: 'NoneType' object has no attribute 'lowvram' and

RuntimeError: "addmm_impl_cpu_" not implemented for 'Half'

(Among the rest of the code)

[karmic crown]

show me the start of log from console

the part where it shows python version and params

################################################################
Launching launch.py...
################################################################
Python 3.10.14 (main, Mar 19 2024, 21:46:16) [Clang 15.0.0 (clang-1500.3.9.4)]
Version: v1.9.0
Commit hash: adadb4e3c7382bf3e4f7519126cd6c70f4f8557b
Launching Web UI with arguments: --skip-torch-cuda-test --opt-sub-quad-attention --upcast-sampling --no-half-vae --use-cpu interrogate --no-download-sd-model

this part

[unique bluff]

Version: v1.9.0
Commit hash: adadb4e3c7382bf3e4f7519126cd6c70f4f8557b
ControlNet init warning: Unable to install insightface automatically. Please try run `pip install insightface` manually.
Launching Web UI with arguments: --skip-torch-cuda-test --opt-sub-quad-attention --upcast-sampling --no-half-vae --use-cpu interrogate --no-download-sd-model
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
Warning: caught exception 'Torch not compiled with CUDA enabled', memory monitor disabled
ControlNet preprocessor location: /Users/sweet.ckae/stable-diffusion-webui/extensions/sd-webui-controlnet/annotator/downloads
2024-04-21 16:35:05,209 - ControlNet - INFO - ControlNet v1.1.443
2024-04-21 16:35:05,495 - ControlNet - INFO - ControlNet v1.1.443
Loading weights [876b4c7ba5] from /Users/sweet.cake/stable-diffusion-webui/models/Stable-diffusion/cetusMix_Whalefall2.safetensors
2024-04-21 16:35:08,248 - ControlNet - INFO - ControlNet UI callback registered.
Running on local URL:  http://127.0.0.1:7860```

[karmic crown]

ok, lets fix this first

ControlNet init warning: Unable to install insightface automatically. Please try run pip install insightface manually.

give me a second

venv/bin/pip install protobuf==3.20.3
venv/bin/pip install insightface==0.7.3

then run it again, and that error should be gone

you might need to add --no-half 😦

but lets do one by one

[unique bluff]

One sec

[unique bluff]

It didn’t work

My computer just loves me I guess

[karmic crown]

run one by one

first
venv/bin/pip install protobuf==3.20.3
then
venv/bin/pip install insightface==0.7.3

[unique bluff]

Done

It had just done the first command

[karmic crown]

and second?

[unique bluff]

Yes

Once I ran the commands separately it ran both

[karmic crown]

good

[unique bluff]

Ill open webui again 🤔

[karmic crown]

run webui again and this part should be gone
ControlNet init warning: Unable to install insightface automatically. Please try run pip install insightface manually.

[unique bluff]

It is gone but…

[north tundra]

Finally fix it, I just reinstalled the a1111

[unique bluff]

Version: v1.9.0
Commit hash: adadb4e3c7382bf3e4f7519126cd6c70f4f8557b
Launching Web UI with arguments: --skip-torch-cuda-test --opt-sub-quad-attention --upcast-sampling --no-half-vae --use-cpu interrogate --no-download-sd-model
Traceback (most recent call last):
  File "/Users/duda.obladen/stable-diffusion-webui/launch.py", line 48, in <module>
    main()
  File "/Users/duda.obladen/stable-diffusion-webui/launch.py", line 44, in main
    start()
  File "/Users/duda.obladen/stable-diffusion-webui/modules/launch_utils.py", line 465, in start
    import webui
  File "/Users/duda.obladen/stable-diffusion-webui/webui.py", line 13, in <module>
    initialize.imports()
  File "/Users/duda.obladen/stable-diffusion-webui/modules/initialize.py", line 23, in imports
    import gradio  # noqa: F401
  File "/Users/duda.obladen/stable-diffusion-webui/venv/lib/python3.10/site-packages/gradio/__init__.py", line 3, in <module>
    import gradio.components as components
  File "/Users/duda.obladen/stable-diffusion-webui/venv/lib/python3.10/site-packages/gradio/components/__init__.py", line 1, in <module>
    from gradio.components.annotated_image import AnnotatedImage
  File "/Users/duda.obladen/stable-diffusion-webui/venv/lib/python3.10/site-packages/gradio/components/annotated_image.py", line 13, in <module>
    from gradio.components.base import IOComponent, _Keywords```

    from fastapi import UploadFile
  File "/Users/duda.obladen/stable-diffusion-webui/venv/lib/python3.10/site-packages/fastapi/__init__.py", line 7, in <module>
    from .applications import FastAPI as FastAPI
  File "/Users/duda.obladen/stable-diffusion-webui/venv/lib/python3.10/site-packages/fastapi/applications.py", line 15, in <module>
    from fastapi import routing
  File "/Users/duda.obladen/stable-diffusion-webui/venv/lib/python3.10/site-packages/fastapi/routing.py", line 22, in <module>
    from fastapi import params
  File "/Users/duda.obladen/stable-diffusion-webui/venv/lib/python3.10/site-packages/fastapi/params.py", line 4, in <module>
    from pydantic.fields import FieldInfo, Undefined
ImportError: cannot import name 'Undefined' from 'pydantic.fields' (/Users/duda.obladen/stable-diffusion-webui/venv/lib/python3.10/site-packages/pydantic/fields.py)
duda.obladen@Marias-iMac stable-diffusion-webui %```

[karmic crown]

what you have before this

Traceback (most recent call last):
File "/Users/duda.obladen/stable-diffusion-webui/launch.py", line 48, in <module>
main()

[unique bluff]

Python 3.10.14 (main, Mar 20 2024, 03:51:51) [Clang 14.0.0 (clang-1400.0.29.202)]
Version: v1.9.0
Commit hash: adadb4e3c7382bf3e4f7519126cd6c70f4f8557b
Launching Web UI with arguments: --skip-torch-cuda-test --opt-sub-quad-attention --upcast-sampling --no-half-vae --use-cpu interrogate --no-download-sd-model

[karmic crown]

please do
git status
and show me the output

[north tundra]

But now I am having a trouble with models with sdxl and also with 1.5, vae's are correct and I tried different parameters and also tried to change clip skip from 1 to 2. Faces are so bad

[unique bluff]

Tu rama está actualizada con 'origin/master'.

Cambios no rastreados para el commit:
  (usa "git add <archivo>..." para actualizar lo que será confirmado)
  (usa "git restore <archivo>..." para descartar los cambios en el directorio de trabajo)
    modificados:     webui-user.sh

Archivos sin seguimiento:
  (usa "git add <archivo>..." para incluirlo a lo que será confirmado)
    .DS_Store

sin cambios agregados al commit (usa "git add" y/o "git commit -a")
duda.obladen@Marias-iMac stable-diffusion-webui % ```

[karmic crown]

looks fine

hmm

replace this

is there any diffrence?

[unique bluff]

Okay, let me do that

[karmic crown]

have you tried forge? it might work better for you, since you only have 8GB of RAM

[unique bluff]

Still made an error

[unique bluff]

How do I do that? I didn’t try it yet

if it helps then great

[karmic crown]

https://github.com/viking1304/a1111-setup/releases

download my script and run it like this

bash a1111-setup.sh -o forge -f all

it will install forge in stable-diffusion-webui-forge folder inside your home folder

it will not affect your a1111 install

[unique bluff]

I need to remove the current one, correct?

Oh

[karmic crown]

no

[unique bluff]

Cool

[karmic crown]

you can have both

almost forgot

forge has controlnet alredy installed

do not try to install it

also

you can use models from a1111

you just need to add --forge-ref-a1111-home /Users/duda.obladen/stable-diffusion-webuiat the end of your params in webui-user.sh in forge folder

i hope i properly copied your username and folder from your log

i will check later what you did, but now i really need to go

[ornate elk]

try with an other model

[gritty ridge]

do we need forge necessary- i installed SD on my M3 Max Pro with 40 core GPU and SD locally installed works fine ... (kohya_ss and dreambooth not at all ) another question- any good model for inpainting for architects to recommend ? i found out i can use only a few models... I need depth and inpainting .... combine them with anyLora...

[unique bluff]

(Yeah, the forge install didn’t work)

Sadge

It didnt even open

[north tundra]

Tried with 3-4 different model but nothing changed

[ornate elk]

do you have face restoration on, if so, disable it

and then install the !After Detailer extension, that will fix faces

[north tundra]

I don't have it and as you can see it's not only about the face :/ I do hires fix for this one and nothing changed

[ornate elk]

can you show your settings?

[karmic crown]

what error you got when you tried to open it?

[karmic crown]

for Macs more RAM can be more useful than more cores. how much RAM do you have? 18, 36 or more?
I have 18GB and my M3 Pro works just slightly faster with forge. with default params a1111 works faster

[karmic crown]

try with 1024x768 and without hires fix, do you get a normal image in that case?

if you do, try with same dimensions, but this time with hires

also try other sampler, some models does not work correctly with some samplers

[ornate elk]

You need to use some quality tags like, photo of people ..., photorealistic, 4k, best quality,
and some negative tags like:
blurry, deformed, mutated, ugly,

[quick forge]

I have insightface installed in A1111.....but I don't see a tab or anything for it anywhere.

[karmic crown]

that is library that controlnet use, there is no tab for that

[quick forge]

ah ok I couldn't remember if it did or not.

are there any checkpoints for doing stuff like generating logos and designs for tshirts and print on demand stuff?

stickers and such

[ornate elk]

Does somebody know if CosXL works in auto1111?

[gritty ridge]

128GB

[fiery knoll]

hello there is page on the server i cant find it tell how to actualy use the ai and it had links to an online website to use it but i cant find it can someone help

[past willow]

Hey guys, I've downloaded SDXL 1.0 loras into the loras folder in A1111, but whenever I generate images with added lora tags and keywords, the end result is the same as without the loras activated, what could be a problem?

Loras are recognized by A1111 in the loras tab.

[vapid jacinth]

ive got 11.2Gb of vram on my gpu. Im trying to get stable diff xl to work on my computer, but keep getting the NoneType object lowvram error. I read somewhere that it should use 12gb of vram and there are plenty of ways to work around it using less. anybody know what those ways are? (note: i did search this first in this channel and saw other people having same issue but havent really found a solution yet....) im on windows 10.

[thorny iron]

Can Stable Diffusion do Image to Video

[mellow wagon]

Just a quick question, for animatediff with Automatic1111, do you need controlnet to use the API?

[drowsy summit]

saw this solution

not letting me edit webui-user.sh

[sour gulch]

Trying to recreate this image from CivitAI: https://civitai.com/images/1519380

Dragged and dropped the image into Comfy, but I get a different result that looks worse.

[narrow copper]

Ahoy. When I'm using comfy ui, and I've got embeddings in their proper folder. How do I activate them? Just put the trigger words in the prompts?

[thorny iron]

I want to learn ComfyUI to make Image to Videos, know a good tutorial of where to start? Does Stable Diffusion have it itself?

[narrow copper]

No idea, sorry, I'm only using text to image.

[ornate elk]

Hey make sure the webui is whitelisted in any browser adblocker.
Then go into models/stable-diffusion folder.
There delete the 1.5 EMA pruned model.
Then download the Dreamshaper v8 model and place it in there.
Then relaunch the webui-user.bat.
https://civitai.com/models/4384/dreamshaper

[drowsy summit]

will try thank u

[ornate elk]

Np, that lowvram error is a known issue since a few days, i still try to figure out what's causing it

[drowsy summit]

oh should i switch to lowvram instead of medvram

[chilly osprey]

Is anyone familiar with Jinja? I'm trying to use the Dynamic Prompts template and it is not working

[ornate elk]

Yes, needed for 4gb vram

[drowsy summit]

ok that worked thank you

i tried to generate a cat and right before it finished it gave me this error

[ornate elk]

What's in your webui-user.bat?

[drowsy summit]

wait im trying to generate again

[ornate elk]

Make sure you have only
--lowvram --opt-sub-quad-attention --opt-split-attention --upcast-sampling --no-half-vae in there.
--xformers is for nvidia only

[drowsy summit]

ok i took out xformers

same thing happened where it gets to around 80% and errors me

heres my webui

[ornate elk]

You need to delete the venv folder and relaunch. Xformers stuff is still in there. It will get removed when deleting the venv

[vocal burrow]

directml ? not zluda ?

[ornate elk]

He has a APU

[vocal burrow]

rip

[fallen quest]

F

[north tundra]

Got it but it was not like this before. I never generated this kind of bad images even if I write bad prompt and without writing negatives. Something is really wrong with my webui but not sure what it is

[north tundra]

I use recommended samplers for each model

I am pretty sure it's not about the parameters

Should I try deleting venv folder or re-install the whole repo?

[ornate elk]

You can try delete the ui-config.json and the config.json for a complete settings reset.
If that doesn't help you can delete the venv folder too.

[hot olive]

how fast is the rx 6500 xt to generate simple images?

[vast crest]

Hi, I noticed that sometimes during the creation of a sequence, the first photos create beautiful and colorful photos, then as I progress they become very dark and I don't know why. I'll leave you some photos so you can understand each other

[ornate elk]

Pretty slow and very much limited of its 4gb vram

[ornate elk]

Hey, what do you mean with "progress"

These are two different images.

[hot olive]

does anyone know how to run SD on colab? and is it still supported?

[vast crest]

no, it's the same image. It's just that as the time sequence passes, in addition to making the images dark, the character's appearance also begins to change.

I'll try to send more photos, I'll be right back

[ornate elk]

You need to explain it a bit more.
Is the first one a preview image? And the second one the output?

[ornate elk]

Ahhh Loopback Wavescript

Such information would help in the initial question xD

[vast crest]

I forgot to mention it XDDD

[hot olive]

this is for RX 6500 xt?

[ornate elk]

Your GPU would need:
--use-directml --lowvram --opt-sub-quad-attention --opt-split-attention --no-half-vae --upcast-sampling

And then you have to use 1.5 models. These are 2gb in size

[hot olive]

--skip-torch-cuda-test ?

[ornate elk]

Nope, that would use CPU

[hot olive]

arg, thanks i will try