🤝｜tech-support

[ornate elk]

You can checkout my AMD directml install guide in the pinned messages of this channel. (Not zluda)

[hot olive]

can you tag the messages bro

[ornate elk]

Sure, here:
#🤝｜tech-support message

You need to lower the additional denois strength

[karmic crown]

you can use the same params i do for a1111:

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

--use-cpu interrogate might be useless for us, but i keep it since it does not have any negative impact

[unique bluff]

03:51:51) [Clang
14.0.0 (clang-1400.0.29.202) 
Version: f0.0.17v1.8.0rc-latest-276-g29be1da7
Commit hash: 29belda7cf2b5dccfc70fbdd33eb35c56a31ffb7
Legacy Preprocessor init warning: Unable to install insightface automatically. P lease try run 'pip install insightface' manually.
Launching Web UI with arguments: --skip-torch-cuda-test --attention-pytorch --al
1-in-fp16 -always-high-vram --use-cpu interrogate
Traceback (most recent call last):
File "/Users/duda.obladen/stable-diffusion-webui-forge/launch.py", line 51, in ‹module>
main()
main
File "/Users/duda.obladen/stable-diffusion-webui-forge/launch-py", line 47, in
start)
File "/Users/duda.obladen/stable-diffusion-webui-forge/modules/launch_utils.py
start ( )
File "/Users/duda. obladen/stable-diffusion-webui-forge/modules/launch_utils.py
", line 541, in start import webui
File "/Users/duda.obladen/stable-diffusion-webui-forge/webui.py", line 17, in <module>
initialize_forge()
File "/Users/duda. obladen/stable-diffusion-webui-forge/modules_forge/initializ ation-py", line 50, in initialize_forge
import 1dm_patched.modules.model_management as model_management
File "/Users/duda.obladen/stable-diffusion-webui-forge/ldm_patched/modules/mod el_management-py", line 122, in ‹module>
total_vram = get_total_memory (get_torch_device()) / (1024 * 1024)
File "/Users/duda.obladen/stable-diffusion-webui-forge/ldm_patched/modules/mod el_management•py", line 91, in get_torch_device
return torch.device(torch.cuda.current_device())
File "/Users/duda.obladen/stable-diffusion-webui-forge/venv/lib/python3.10/sit e-packages/torch/cuda/__init__.py", line 803, in current_device
-lazy_init()
File "/Users/duda obladen/stable-diffusion-webui-forge/venv/lib/python3.10/sit e-packages/torch/cuda/__init__py", line 309, in _lazy_init
raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enabled duda.obladen@Marias-iMac stable-diffusion-webui-forge % |```

This was the (almost) whole code

[karmic crown]

strange

Legacy Preprocessor init warning: Unable to install insightface automatically. P lease try run 'pip install insightface' manually.

you get this after using my script?

do this thing again, but this time in that folder:

first
venv/bin/pip install protobuf==3.20.3

then
venv/bin/pip install insightface==0.7.3

it should work after that

[dense tapir]

Spam

@mods

[solemn moon]

TypeError: unsupported operand type(s) for -: 'tuple' and 'int'

bruh

[noble belfry]

you need to give way more context than that.

[solemn moon]

ahh, could be nvidea driver update

[noble belfry]

in general if you see something saying it doesnt know what a module is, is an extension assuming something is there but it isnt. nearly always fixed just by updating things

[grim moon]

hey everyone, just booted sd and this pops up venv "C:\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 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)] Version: v1.9.0-6-gb7c5e603 Commit hash: b7c5e6039e3a60dfd0b4df342782fabdf6844a16 Traceback (most recent call last): File "C:\SD-Zluda\stable-diffusion-webui-directml\launch.py", line 48, in <module> main() File "C:\SD-Zluda\stable-diffusion-webui-directml\launch.py", line 39, in main prepare_environment() File "C:\SD-Zluda\stable-diffusion-webui-directml\modules\launch_utils.py", line 593, 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 Press any key to continue . . .

Any ideas? when i add --skip-torch-cuda-test it gave me another error

It worked for me yesterday just fine

[solemn moon]

Did you update yr drivers?

[grim moon]

if you mean gpu - yes

[solemn moon]

working again, thanks for the lot 🫡

[noble belfry]

well you sorted out your own problem, but just in future so you know. they're called upstream issues and the bane of github projects. the more you know 🙂

[grim moon]

Any ideas? I installed it yesterday. It worked yesterday too

i installed the amd driver again and it works

Which is in fact weird. bc i installed the same file yesterday

[nova wren]

is it me or word combinations with AND just breaks all semblance of loras?

has anyone tested if keyword combinations with {|} can be made to somehow work with lora?

[ornate elk]

remove --skip-torch...

[maiden coral]

Hello everyone. I'm new to stable diffusion. I installed AUTOMATIC111 diffusion from github and all dependencies (Python 3.11.9). It worked as usual until today.
The image with more than 20 steps cannot be generated and I noticed that my RAM and VRAM just overflows (without any error message because the program either crashes or in cmd it shows "press any key to continue..." without any error message.

• I have Intel core i5 10300H, GTX 1650 4 GB VRAM and 2x8 GB DDR4
  I usually generated 5 batches of images 640x640 or 512x512 and then upscaled them. Now I cannot even generated them because the program crashes and system lags too with low amount of RAM.

My parameters in webui-user.bat (without low-vram it would display the CUDA overflow error):

• set COMMANDLINE_ARGS=--no-half --opt-channelslast --upcast-sampling --lowvram --gradio-img2img-tool color-sketch (previously used --xformers, but that didn't help the issue)
• set PYTORCH_CUDA_ALLOC_CONF=garbage_collection_threshold:0.9,max_split_size_mb:512
  Solution I tried: I uninstalled every dependency and then reinstalled everything with the newest version. **I'm also using Python 3.11.9 ** which works too, but today it stopped working for unknown reasons. Creating a simple prompt works but takes a lot of time. Also tried to set a global variable Pytorch_cuda... to 128MB or 512MB and still didn't work. I was using it yesterday and could do other things on the notebook, now I can't.
  My GPU is working as usual. I haven't found any solution or leads for this problem. I'll be glad for any input that's resolve it. Thank you.

[maiden coral]

Which program for stable diffusion are you using? Install the prerequisites from requirements.txt file with "pip install -r requirements_file.txt" and disable hardware acceleration in Windows in order to show "CUDA" option in taskbar. This worked for me. I use AUTOMATIC111, although I have a problem too, but it might work for you.

[sinful needle]

anyone else having issues with those chinese icon packs? https://civitai.com/models/47800/game-icon-institutemode , https://civitai.com/models/143301/game-icon-instituteanfgv3

they aren't working properly for me. not sure what im doing wrong

[grim moon]

i was using sd-zelda, reinstalled amd dirvers, and it worked

[maiden coral]

So drivers were the issue, have you ever had a problem with overflowing RAM/VRAM?

it all started today ...

[maiden coral]

Have you checked the trigger words? Are they contained in the first line if you open the model with Notepad++?

[sinful needle]

lemme see. translated the page and they said they only require "Game Icon Research Institute" as the trigger

im trying to achieve a stone tablet and it keeps putting anime girls on it. their examples dont have that problem

[maiden coral]

because in my opinion there are many icons and every single one has a specific trigger

[sinful needle]

ill check hol up

[maiden coral]

give negative prompt (((anime girl))) and these words, also what model are you using?

[sinful needle]

opened it in notepat ++ and this is the ONLY info: "description": "3D Icons by Game Icon Institute. ",
"notes": "1.5F_Ico_Digital__GameInstitute3D_--Semi\nConceptConnoisseur | Game Icon Institute3d, Euler a, 7, 20, 2, 512x512\nhttps://civitai.com/models/284567",
"vae": "1.5_Style_AyoniMix.vae.ckpt"
}

[maiden coral]

studied these brackets yesterday, [] for the things you don't want as positive, but it can contain only one phrase such as [anime girl] in positive prompt

lemme see, some models don't have the train words, I'll download it

[sinful needle]

im using both the model AND the lora i linked

the lora is closer to the aesthetic i want since im making fantasy themed icons for desktop

thanks

[maiden coral]

it'll take 10 min. to download it, let me check, also the line should be longer

[maiden coral]

check the first line, double click and open it in MS word or google translate, then look for inverted backslashes where the words are contained

\ and try encoding in JP or CN language if it isn't written in English, that's the only thing I cna thing of

[sinful needle]

Not an icon at all:

Positive Prompt: (stone square tablet, gray:1.2), lora:1.5_Ico_Digital_-_GameInstitute_Fantasy--unreal:0.06 Game Icon Research Institute

Negative: (Anime girl, woman, human, male, person:1.5). ((1.5_Neg_Easy:0.8), black, dark, Low resolution,:1)

ignore the double underscores, which is negligable and tested. I have a specific sorting method for my LoRA etc.

[maiden coral]

I wouldn't use underscores, but it might be the embedding, right, such as bad_hands-5 and so on

[sinful needle]

Usually I'd ignore it, but some models come with underscores in it already and it has never impeded my output. No issues til now. Tried without it too just in case and to weed out possibilities

[maiden coral]

so this is lora model but a normal model, I see

[sinful needle]

im using both their large model and a linked lora model

the loras guide to specific styles whereas the large checkpoint is to make icons in general

[ornate elk]

Hey your first issue is python 3.11
Its not supported by Stable diffusion and can cause unknown issues. You need 3.10.11 and delete the venv folder after installing it.
The next thing is your webui-user.bat commandline_args
You need to remove --lowvram and instead add --medvram --xformers
And remove Py torch cuda alloc garbage collection split size completely
Then relaunch the webui-user.bat

[sinful needle]

i went to their main GameResearchInstitute discord but its a ghost town. fishy

[maiden coral]

Thank you, I'll try it. It was working just fine yesterday with 3.11.9 version of Python. Idk what went wrong.

[sinful needle]

adding "white background" SLIGHTLY helped with generating icons, but it's still not matching the fantasy theme of the LORA, just the main checkpoint. too lighthearted

[maiden coral]

check the bilibili video, although I can't see clearly because it's the quality video I can see:

• prompt: image is an apple, blue, ... I can't read it, game icon, official art, well-structured, HD, 2d, game prop icon, {white background}, linked lora*
• negative prompt: (easynegative:0.8 or something like that), black, dark, Low resolution

[sinful needle]

roger roger ill try thanks

[maiden coral]

so try a similar prompt but specify an object you want to try

[karmic crown]

3.11.x should work, but it is not officially supported, so it's better to use 3.10.x

[maiden coral]

https://www.bilibili.com/video/BV1LX4y1v77z/?share_source=copy_web&vd_source=34a51e5ffd9c5724f30396393ab75ef6

[maiden coral]

Ok, what about the requirements? Is it enough to download the newest (which I have now) or download the ones from "requirements_version.txt"?

[karmic crown]

you do not need to bother with that yourself with a1111

just remove venv folder and let a1111 create new one and install everything

[maiden coral]

idk, it doesn't seem complicated except memory overflow, idk if the ocrrect version of python fix these things

ok, I'll try to do so, thank you

[karmic crown]

it is best to follow one of the guides in the pinned messages of this channel

[ornate elk]

Memory overflow will be fixed by using --xformers mostly

But if you gen at to high resolution you also need the tiled VAE extension

[maiden coral]

I'll look it up. I mostly tried to normalize the img generation and tune the generation time down without losing quality.

[ornate elk]

Yea none of the commandline args will alter the quality

[karmic crown]

speed and quality are not related

[sinful needle]

holy crap it worked, something about "official art, well constructed" etc. did it. but for some reason only 1 out of 10 times does it make a suit of armor like instructed. and used a sampler they didnt use. idk why the one they used doesnt work

[maiden coral]

xformers should make performance a bit faster or so I heard

[ornate elk]

A lot faster

While reducing vram.usage

[sinful needle]

i wish there was a way to make the output a certain hexcode in color. i want my background to be a specific shade of green so i can chroma key it out

[maiden coral]

yes, this has been bugging me, I should be able to use it even with 3.11, I know about someone who is using Python 3.11, I'm still looking up things and these plugins that are used for it

[maiden coral]

if you don't know which shade you can "google" them through AI, that's how I do some things

[sinful needle]

but if i were to put the specific hexcode in the prompt id doubt it would use it

ill try

#00b140 <----------- is the specific shade of green

[maiden coral]

https://www.color-name.com/hex/cc00cc you can try to use this

[sinful needle]

thanks

[karmic crown]

i am not saying that 3.11 will not work, i am just saying it is not officially supported
if you have a problem and raise a ticket, it will be ignored, since you are not using the supported version of python

[sinful needle]

your site might be bugged. i type any hex code and press enter and it always turns up with 12C9CC venetian red

[maiden coral]

Green (Pantone), yeah, just don't copy the code into it

you must manually delete it and write it manually too

[sinful needle]

oh copy might be doing something roger roger

[maiden coral]

the website automatically inserts it's own randomly generated colour, it works if you delete it and write it manually

[sinful needle]

ty

getting the thing to work moderatly now. had to try literally every sampler and it turned out to not be the one they used...

[sinful needle]

Game Icon Institute, game icon, ahfg, (1 Open red and black book:1.2), magic swirls, still life, black background lora:1.5_Ico_Digital__GameInstitute_Fantasy--unreal:0.06

[karmic crown]

0.06 <- this looks too low, it is usually something like 0.6

and be sure that this 1.5_Ico_Digital__GameInstitute_Fantasy--unreal matches the filename of lora

better rename it to somthing simpler

[sinful needle]

Yeah i have a weird system. my weaponized autism naming scheme:

version-family(figure, style, etc.)_type_media(photo, digital, etc.)_brand associated with it_actual name of model_genre-tone-how realistic it is

[karmic crown]

as long as this match the filename 1.5_Ico_Digital__GameInstitute_Fantasy--unreal you should be fine

[sinful needle]

oh yeah i do that too. i name the pictures associated with it the same and the .json. nice previews. comfy

much categorize. such tidy

[karmic crown]

but as i said, this is probably too low 0.06

[sinful needle]

for some reason thats what they recommended in all examples. going higher made it freak out

[karmic crown]

do what they do

[sinful needle]

finally turned out well. gonna make some sick fantasy icons to go with my razer axon motion desktop background

[karmic crown]

i am just saying that normal loras use something like 0.6, 0.8...

[sinful needle]

i wanna make a motion background of Siophra river from elden ring wew

[past willow]

Bump

[latent grail]

hi could someone assist me with this? The HR_fix isnt functioning properly. I'm only generating 512x512px images with the api.

[sinful needle]

not sure fren

[maiden coral]

So I installed the exact version of Python and it still doesn't work. It just crashes even with xformers.

[ornate elk]

What are you trying to generate? Which settings are used? Model, resolution etc

[maiden coral]

well, yesterday I tried what are my max specs I can generate. I tried to generate Raiden shogun from Genshin Impact and put the usual prompts. - sampling: 50

• batches: 5
  and 512x512 px or 640x640px resolution and clicked generate

but now it seems it doesn't work anymore, RAM and VRAM overflow -> the app crashes along with Chrome

[soft fiber]

did you delete the venv folder?

[maiden coral]

yes

I did. that's the thing

[ornate elk]

Did you removed the stuff from the webui-user.bat I mentioned?

[maiden coral]

I'll take a look.

Which ones? I think there are no other parameters. I was working with them yesterday.

exactly the same ones

set COMMANDLINE_ARGS=--xformers --no-half --opt-channelslast --upcast-sampling --lowvram --gradio-img2img-tool color-sketch

I only have these now

pytorch argument was doing nothing

[ornate elk]

Yea

You can also remove --opt-channelslast

And what model did you used?

Was it 1.5 or sdxl based?

(2gb or 6gb)

[maiden coral]

"lyuyangMix_v297" this one, SDXL

I haven't had any problems with these models, 6GB I think

I used even 12GB model and no problems

most of them are SD based, but there were no problems with SDXL problems either

[ornate elk]

You also need to make sure Heavy Programms like Wallpaper Engine are not running or games

[maiden coral]

yeah, they are not, the only thing that can eat RAM is excel, chrome, Discord and notepad++

but I should have enough, no reason to use all the RAM and lag the entire computer

[ornate elk]

Make sure you also have more than 8gb RAM and 15gb free space on C drive

Running Sdxl models with a 1650 is a really bad experience and not recommended

[maiden coral]

I do have. and they are both 500GB and 1 TB SSDs in VNMe slots

40 min to generate 640x640 px with sampling: 50 but it was SD based model

5 pictures*

[ornate elk]

Thats crazy why not using 1.5 models that are much smaller and generate faster with your card?
You can then upscale them to any resolution in img2img in less amount of time

[maiden coral]

I mean ... that's normal. I don't have a powerful notebook. I did upscale it to 2560x2560

the result was pretty good, although hands need to be worked on with prompts

any fix for this? It's about comfyUI

[ornate elk]

Thats a python issue

Because you had 3.11

[maiden coral]

I'm using 3.10.6

I uninstalled 3.11

[ornate elk]

Ok

[ornate elk]

Wait it says rocm

Do you have an AMD CPU?

[maiden coral]

nope

[ornate elk]

Okay very strange then

[maiden coral]

oh wait, it's for AMD

yeah, wrong one

but I think it'll be the same. I downloaded requirements before and installed them manually.

[ornate elk]

you dont need to install any requirement manually

comfyui comes with an install script

same goes for auto1111

[maiden coral]

too late for that, I generated one when I had A111

[ornate elk]

then maybe delete comfyui, download it again and just run the comfyui start bat

it should do everything automaticly

[maiden coral]

it still shows the same error

maybe it needs a specific version of torch

I installed this one: "torch-2.2.2-cp310-cp310-win_amd64"

[ornate elk]

the same error ? then you should open up a cmd and type
Pip cache purge

then run comfyui again

[maiden coral]

same error even with purge

[ornate elk]

the same rocm message?

[maiden coral]

I'll reinstall it again, but I doubt it'll work. maybe restart will help

[shrewd shale]

if I want to move my kohya ss or automatic1111 from a drive to another, how can I do it? the UI is linked to the old drive.
it just works with comfyui because it's portable.
should delete the venv? or there is something can I edit in the venv will solve it?

[maiden coral]

I'd just convert from A111. Thhat app isn't working as intended and instead eats so much VRAM and RAM.

[ornate elk]

You can just move the whole folder. But moving it would break the git path behind it. So you may can't update it and need to set the git path again

[ornate elk]

Lol it works as intended but you try to use an sdxl model with a gpu that is not made for that

[maiden coral]

it worked on other models based on SXDL

I found ComfyUI puzzling. Portable version works but not the one on the main page of github.

[ornate elk]

It can work yes, but it doesn't have to work all the time. If you get the tiled VAE extension it could work more often

[ornate elk]

The main comfyui version is the portable version you download as zip

[maiden coral]

well, it was displayed as a portable version, so I thought it was a "side" version.

like an alternative

but this one works, the other one can't works without pytorch converted with CUDA or whatever error that is

I had this one on A111 too, but it was solved with reinstalling, this one wasn't

[ornate elk]

If the portable has also the update.bat inside it then your good

[maiden coral]

yup, thanks, I'll try it, it looks intuitive

[obsidian carbon]

Hey so I had talked here in the past about issues about A1111 and I tried everything but I found out a remedy that actually fixed my end of things regarding my stable diffusion.

For some reason, any updates after SD 1.7.0 seem to cause a memory leak in my ram. Now granted I can’t say that this will work for everyone but I’m gunna link a comment of someone who said what they did to fix the memory leak. Honestly, until SD 3 comes out, here’s the best thing I had to use

Also for reference I’ve 32GB ram so it helps having that as I don’t think this would of helped me if I still had my 20GB

Very bottom, comment from “rococo-p”

https://github.com/AUTOMATIC1111/stable-diffusion-webui/discussions/5094

Also, wanted to ask, would I also get help in this channel regarding Kohya? Or where else should I ask if not here ?

[unique bluff]

Should I reinstall forge sd? It’s not opening (it gives an error)

[sinful needle]

hi im getting this error when using AnimateDiff extension. never got it until just today. Worked fine yesterday:

NansException: A tensor with all NaNs was produced in VAE. Use --disable-nan-check commandline argument to disable this check.
Time taken: 33.3 sec.

Got the error to go away, but now it's just a black screen instead of an actual gif. was working fine yesterday.

[verbal burrow]

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. Can someone explain this to me?

[maiden coral]

Is there any way to safe a project in ComfyUI? It doesn't queue, but I copied all the nodes, but unfortunately it doesn't work from the space I copied it.

[maiden coral]

Which UI are you using?

[verbal burrow]

Automatic1111

[maiden coral]

switch to ComfyUI

I switched from Automatic111 because if stopped working and had enormous overflow. Especially with certain models.

you'd need to add --no-half into parameters as well as --low-vram if you don't have 8+ GB VRAM

and --xformers to optimize it if you have Nvidia GPU

seems like I'll need to start over, this took me quite a lot of time to set up

[sinful needle]

When trying to use AnimateDiff:

RuntimeError: CUDA error: device-side assert triggered
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1.
Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.

[maiden coral]

idk what that is, I assume some kind of plugin that cna make the images animated

What is your GPU?

I've also got an error while scaling an image. "CUDA out of memory. Tried to allocate 258.00 MiB. GPU 0 has a total capacity of 4.00 GiB of which 0 bytes is free. Of the allocated memory 3.16 GiB is allocated by PyTorch, and 105.39 MiB 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)" What are the icommon commands to fix this issue for ComfyUI?

[sinful needle]

actually ive narrowed it down to that it doesnt recognize the extension/module:

File "C:\Users\place\Desktop\Studio\Visual\Image\Ai\Stable Diffusion\stable-diffusion-webui\extensions\sd-webui-animatediff\scripts\animatediff.py", line 43, in ui
from scripts.animatediff_mm import mm_animatediff as motion_module
ModuleNotFoundError: No module named 'scripts.animatediff_mm'

[sonic pine]

What do I have to reinstall here? I tried to add xformers but its all messed up now and wont launch

Python 3.9.13 (tags/v3.9.13:6de2ca5, May 17 2022, 16:36:42) [MSC v.1929 64 bit (AMD64)]
Version: v1.8.0
Commit hash: bef51aed032c0aaa5cfd80445bc4cf0d85b408b5
Installing xformers
Launching Web UI with arguments: --xformers --reinstall-xformers
WARNING:xformers:WARNING[XFORMERS]: xFormers can't load C++/CUDA extensions. xFormers was built for:
PyTorch 2.1.2+cu121 with CUDA 1201 (you have 1.13.1+cu117)
Python 3.9.13 (you have 3.9.13)
Please reinstall xformers (see h ttps://github.com/facebookresearch/xformers#installing-xformers)
Memory-efficient attention, SwiGLU, sparse and more won't be available.
Set XFORMERS_MORE_DETAILS=1 for more details
No SDP backend available, likely because you are running in pytorch versions < 2.0. In fact, you are using PyTorch 1.13.1+cu117. You might want to consider upgrading.
Traceback (most recent call last):
File "Z:\Other\stable-diffusion-webui\launch.py", line 48, in <module>
main()
File "Z:\Other\stable-diffusion-webui\launch.py", line 44, in main
start()
File "Z:\Other\stable-diffusion-webui\modules\launch_utils.py", line 465, in start
import webui
File "Z:\Other\stable-diffusion-webui\webui.py", line 13, in <module>
initialize.imports()
File "Z:\Other\stable-diffusion-webui\modules\initialize.py", line 36, in imports
shared_init.initialize()
File "Z:\Other\stable-diffusion-webui\modules\shared_init.py", line 40, in initialize
from modules import styles
File "Z:\Other\stable-diffusion-webui\modules\styles.py", line 9, in <module>
class PromptStyle(typing.NamedTuple):
File "Z:\Other\stable-diffusion-webui\modules\styles.py", line 11, in PromptStyle
prompt: str | None
TypeError: unsupported operand type(s) for |: 'type' and 'NoneType'
Press any key to continue . . .

[sinful needle]

How do i upgrade Pip? I opened python and entered "-m pip install --upgrade pip" but i get an invalid syntax error

[solemn moon]

try deleting venv folder then rerun

[solemn moon]

did you upgrade yr gpu software

[sinful needle]

i use geforce experience to update my gpu if thats what u mean i guess

unless im missing something more

[solemn moon]

yea

[sinful needle]

yeah

[solemn moon]

try delete venv folder

[sinful needle]

already deleted venv

[solemn moon]

my go to nsolutiom

[sinful needle]

and reran

[solemn moon]

hmm

restart pc?

[ornate elk]

What's your GPU?

[sinful needle]

tried upgrading pip thru cmd and it was finding a wrong path or something

[verbal burrow]

RTX 3060

[solemn moon]

yea ive had this b4 trying to remember what happened though

[sinful needle]

also my main issue right now is animatediff:
File "C:\Users\place\Desktop\Studio\Visual\Image\Ai\Stable Diffusion\stable-diffusion-webui\extensions\sd-webui-animatediff\scripts\animatediff.py", line 43, in ui
from scripts.animatediff_mm import mm_animatediff as motion_module
ModuleNotFoundError: No module named 'scripts.animatediff_mm'

its weird it was running fine then today WAMMO

just bricked itself

[solemn moon]

maybe something to do with animatediff

[sinful needle]

the whole automatic1111 craps itself unless i delete the extension

[solemn moon]

like delte that extension or summin

[ornate elk]

Okay and when do you get that error?

[sinful needle]

no the pip thing was going on for a long time i just ignored it

well i kinda wanna use the extension to animate

[solemn moon]

bruh

[sinful needle]

thats my pickle

[verbal burrow]

When pressing generate

[sinful needle]

just confused how it worked one day then bricks the entirety of automatic1111 the next

[ornate elk]

Okay, then check the model you trying to use.
Make sure its 2gb or larger

[verbal burrow]

It's 6 GB

[ornate elk]

Okay so its sdxl

[verbal burrow]

I was trying to use a character sheet from civitai that said it's supposed to be used with Pony Diffusion

[solemn moon]

are you sure yr running right launch file. webui-user.bat

[verbal burrow]

It's labeled as Lora Model

[sinful needle]

yep. always

ive been using automatic1111 for a year the same way

[ornate elk]

Okay, then make sure the lora is in the models/lora folder.
And pony v6 is in models/stable-diffusion

[verbal burrow]

https://civitai.com/models/368139/character-sheet

This one

[sinful needle]

well i just found out how to install pip. that parts fixed

"python -m pip install --upgrade pip" in cmd

not actually running it in python shell

[solemn moon]

[sinful needle]

now just need animatediff to stop crapping itself so i can make anims. i wanna make psychadelic wallpapers

i like elden ring. wanna make animated siofra river wallpaper

theres the full error

something to do with launch.py at first then referencing missing modules from animatediff, but the module is indeed there

might just do a clean install

[sonic pine]

any help please

[karmic crown]

This is your problem:

No SDP backend available, likely because you are running in pytorch versions < 2.0. In fact, you are using PyTorch 1.13.1+cu117. You might want to consider upgrading.

[sonic pine]

do you know how to upgrade it? I think I tried before but it didnt change on the screen

[karmic crown]

update a1111 to latest version, remove venv and let a1111 do everything for you

[sonic pine]

gotcha ill try that

should there be a file I need to run to update it?

or do I need to search it up

[karmic crown]

open cmd in that folder and just do
git pull

[sonic pine]

done

[karmic crown]

remove venv folder
and run a1111 as usual

it will detect that venv is missing and it will install what you need

[sonic pine]

yup its downloadin stuff now

ill get back to you once its done

[ornate elk]

Yes thats a lora and goes into models/lora folder

[sinful needle]

does anyone else here use animatediff on automatic1111? anyone else having issues with it?

[sonic pine]

it works now. Thanks for the help!

[ornate elk]

I'm using it with no issues

[sonic pine]

File "Z:\Other\stable-diffusion-webui\venv\lib\site-packages\gradio\routes.py", line 488, in run_predict
output = await app.get_blocks().process_api(
File "Z:\Other\stable-diffusion-webui\venv\lib\site-packages\gradio\blocks.py", line 1434, in process_api
data = self.postprocess_data(fn_index, result["prediction"], state)
File "Z:\Other\stable-diffusion-webui\venv\lib\site-packages\gradio\blocks.py", line 1335, in postprocess_data
prediction_value = block.postprocess(prediction_value)
File "Z:\Other\stable-diffusion-webui\venv\lib\site-packages\gradio\components\gallery.py", line 197, in postprocess
file_path = str(utils.abspath(file))
File "Z:\Other\stable-diffusion-webui\venv\lib\site-packages\gradio\utils.py", line 938, in abspath
if is_symlink or path == path.resolve(): # in case path couldn't be resolved
File "C:\Users\User\miniconda3\lib\pathlib.py", line 1205, in resolve
s = self._flavour.resolve(self, strict=strict)
File "C:\Users\User\miniconda3\lib\pathlib.py", line 206, in resolve
s = self._ext_to_normal(_getfinalpathname(s))
OSError: [WinError 123] The filename, directory name, or volume label syntax is incorrect: 'outputs\txt2img-images\2024-04-22\00000-3461562990.png?1713854338.9990487'

@karmic crown do you know what this line means?

OSError: [WinError 123] The filename, directory name, or volume label syntax is incorrect: 'outputs\txt2img-images\2024-04-22\00000-3461562990.png?1713854338.9990487'

it does output to that directory

idk why its giving me that error

[maiden coral]

What are you trying to do? Is t A111? Have you changed anything recently?

[sonic pine]

nope

just updated it

[maiden coral]

What did you update?

install the recommended version.

"requirements_version.txt" or let the automatically install all dependencies

[mint kayak]

AttributeError: 'UiControlNetUnit' object has no attribute 'enable' .ControlNet v1.1.445 Does anyone have an error?🤔

[maiden coral]

Nope, also what UI you are using. What is you GPU and what are you trying to do?

[mint kayak]

I have ControlNet update to v1.1.445 and since then I have been getting errors. I have now deleted I have ControlNet and reinstalled it and now it seems to be working again 😅

[maiden coral]

I haven't got that far so I can't help you. But try to downgrade for now and also downgrade dependencies if you updated any of them.

and go according to a guide

[mint kayak]

Thank you for wanting to help, it seems like the problem sorted itself out 🥳

[dapper frigate]

Did you find this out?

[dapper frigate]

Ok just tested, it's blob, it would have been better if they didn't use ANY type in documentation

[sweet zenith]

lol yeah

  async postStabilityAiSd3ImageToImage(id: string, imageUrl: string) {
    try {
      let imageBuffer: any = await this.getImageBufferFromUrl(imageUrl);
      imageBuffer = await this.formatForStabilityAi(imageBuffer);
      const formData = new FormData();
      formData.append('image', imageBuffer, { filename: 'image.png' });
      formData.append('mode', 'image-to-image');
      formData.append('model', 'sd3');
      formData.append('strength', 0.65);
      formData.append('prompt', '_');
      formData.append('output_format', 'png');

      const response = await this.sd3GenRequest(formData);

      if (response.status !== 200) {
        console.log(response.data);
        throw new Error(
          `Failed to post to StabilityAi SD3 image to image: ${response.status}`,
        );
      }

      return await this.processSD3Response(response, id);
    } catch (error) {
      throw new Error(
        `Failed to post to StabilityAi SD3 image to image: ${error}`,
      );
    }
  }

async getImageBufferFromUrl(url: string): Promise<ArrayBuffer> {
    try {
      const response = await firstValueFrom(
        this.httpService
          .get(url, {
            responseType: 'arraybuffer',
          })
          .pipe(
            catchError((error: any) => {
              throw new Error(`Failed to get image from url: ${error.message}`);
            }),
          ),
      );
      return response.data;
    } catch (error) {
      throw new Error(`Failed to get image from url: ${error.message}`);
    }
  }

  async sd3GenRequest(formData: FormData): Promise<AxiosResponse<any, any>> {
    const url = `https://api.stability.ai/v2beta/stable-image/generate/sd3`;

    const response = await firstValueFrom(
      this.httpService
        .post(url, formData, {
          validateStatus: undefined,
          responseType: 'arraybuffer',
          headers: {
            ...formData.getHeaders(),
            Authorization: `Bearer ${this.stabilityAiConfig.apiKey}`,
            Accept: 'image/*',
          },
        })
        .pipe(
          catchError((error: any) => {
            throw new Error(
              `Failed to post to StabilityAi SD3 text to image: ${error.response.data.message}`,
            );
          }),
        ),
    );
    return response;
  }

[sweet zenith]

There's my code snippet in case it helps anyone, adding the {filename: 'image.png'}
to the image param of the formdata is what helped me stop getting 400 "image is required" error

[dapper frigate]

Great, thanks for sharing

[maiden coral]

Does anyone know how to limit this error message (limit pytorch in VRAM consumption):
"Error occurred when executing UltimateSDUpscale:
CUDA out of memory. Tried to allocate 128.00 MiB. GPU 0 has a total capacity of 4.00 GiB of which 0 bytes is free. Of the allocated memory 3.38 GiB is allocated by PyTorch, and 26.11 MiB 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)"

[ornate elk]

That message always appears when you don't have enough vram to make a specific task

[maiden coral]

yes, is there a way to limit pytorch?

[ornate elk]

In that case the upscaling failed. Is that on auto1111 or comfyui?

[maiden coral]

comfyui

[ornate elk]

Nope

[maiden coral]

I tried that command and even the variant with garbage collector:

• "set PYTORCH_CUDA_ALLOC_CONF=garbage_collection_threshold:0.6,max_split_size_mb:128" at the beginning of run_nvidia_gpu.bat file and it does nothing

I could upscale it on A111 just fine. I even set tiled_decode=True, still didn't work

I'll try not to do these operations one after another but do them separately.

Maybe a way not to run out of memory while upscaling a picture.

it didn't work either

[hardy canopy]

Can you tell me how to download the LORA LCM model from Huggingface? Apologies for the noob question. I tried downloading the pytorch_lora_weights.safetensors from the 'Files and Versions' tab, but that throws an error in Comfy. "Prompt outputs failed validation. LoraLoader: - Required input is missing: clip"

[maiden coral]

where is the model you need

[hardy canopy]

https://huggingface.co/latent-consistency/lcm-lora-sdv1-5/tree/main

[maiden coral]

Do you have everything setup?

CLIP is a must

[maiden coral]

https://huggingface.co/latent-consistency/lcm-lora-sdv1-5/blob/f8d4aef9472b88fcce04f6dad7567401d70efc91/README.md here you go

readme file

[hardy canopy]

I guess I'm not sure ... ? I'm an A1111 guy, and have had Comfy installed for months, but never use it. I've generated things in the past with Comfy

[maiden coral]

well, comfy is better, but I still get out of memory error because of my low VRAM

[hardy canopy]

Gotcha. So, I'll need to find this mysterious 'CLIP' you speak of?

[fallen quest]

Connect a clip output to the node input that throws the error

Clip comes from the checkpoint loader

[hardy canopy]

dear lord. how do people use comfy. haha. hold please while i go get a doctorate. but seriously, thank you for that

so, here's the workflow i was attempting to use. i wanted to use this specific LCM lora. so, are you saying i need to connect the Lora's CLIP to something here?

[fallen quest]

Pull a connector from the load checkpoint clip to the clip of the lora

Then from the load Lora to the clip text encode

[hardy canopy]

ok. i so i pulled a connector from the load checkpoint clip to the clip of the lora. and that allowed me to queue a prompt

there seems to be 2 CLIP text encode boxes in this workflow. should i connect load lora to both?

[fallen quest]

I only connect the positive

[hardy canopy]

does that look right?

[maiden coral]

Does anyone know what is this "pixels" option I need to match? I'm trying an alternative to ultimate SD upscaler.

found it 👍

[fallen quest]

No. Clip from the Lora loader is still not connected to the clip input of the clip text encode

[maiden coral]

yes, looks good, although idk what is comfyui for photoshop and these options

yes, clip is missing

also, any way to do upscaling to 2560x2560px. I think I've tried everything from the options I havei nstalled.

Error occurred when executing IterativeImageUpscale:

CUDA out of memory. Tried to allocate 288.00 MiB. GPU 0 has a total capacity of 4.00 GiB of which 38.65 MiB is free. Of the allocated memory 3.08 GiB is allocated by PyTorch, and 84.87 MiB 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) but it uses less VRAM (tiled upscaling)

but I'm still getting that error

the consumption was around 3 GB VRAM and then went really high and output this error

Any other way to do upscaling from 640x640 to 2560x2560px without VRAM usage going overboard?

[ornate elk]

SD upscale script in auto1111 or the ultimate upscale extension

Also tiled VAE extension highly recommended

[maiden coral]

thank you, I'll look into these extensions

I tried ultimate SD upscaled with tiled option turned on but it dind't go well

Where is the A111 SD scale script usually stored in ComfyUI? Which folder?

[ornate elk]

You then need to set the tiles to 768x768 and upscale by 2.
If the input resolution is to high you would still get oom but therefore tiled VAE could help

[ornate elk]

The SD upscale script is only in Auto1111 but works the same as the ultimate upscale extension in comfyui

[maiden coral]

it's .py extension not.pth

that's the thing

[ornate elk]

.PTH files are upscalers

The ultimate upscale extension is a script

Which can use upscalers

[maiden coral]

How do I load it? btw it's called: "sd_upscale.py"

[ornate elk]

In comfyui you need the Node for it

[maiden coral]

Which one is it?

[ornate elk]

But with your GPU I would stick to auto1111, as it can use tiled VAE extension and comfyui not

[maiden coral]

well, A111 doesn't work at all

instead it writes connection errored out after using too mcuh VRAM and RAM

and lags computer too

[ornate elk]

Here is one:
https://github.com/ssitu/ComfyUI_UltimateSDUpscale

[maiden coral]

I have this one setup

I tried the pixel tiled one and it still failed with the same message

[uncut jewel]

Anyone having issue with "Import Failed" nodes and know how to solve it ?

[fallen quest]

It uses it by default if the ram is not sufficient

[ornate elk]

If it fails at the last step its mostly cause the VAE gets not tiled and that needs a lot of vram.

[ornate elk]

Can you explain?

[fallen quest]

ComfyUi is switching to tiled vae if it recognises that there is too less Vram for the operation.

[ornate elk]

Ahh okay, that's good to know thx

[lime breach]

Hey Guys!
Is the Membership meant for the commercial use of the model or the output images?

[vocal burrow]

That s a question better suited for their support team / lawyers rather than a bunch of random people on a discord ^^. You can contact them through their website.

(also it depends of the model you re using)

[hardy canopy]

How can I cancel and clear a queue in Comfy?
Cancel and Clear Queue buttons aren't doing anything for me

[heady grove]

Hi I can't run Forge UI how can I fix it

[ornate elk]

Hey, did you modified the run.bat ?

Also what's your python version?

[heady grove]

Hmm this is my run file

No have .bat at behind

3.10

[lime breach]

Ok Thanks!

[ornate elk]

Hmm okay, and whats in RhE webui-user.bat inside the webui folder?

[heady grove]

This?

[maiden coral]

restart comfy (literally end the program and then start it again)

if you closed it or did something then none of the buttons will work

and the progress won't be saved

so if you did anything after that and tried to copy it from the local server that was closed before (session) then it won't work on a new session

A quick question. I'm trying to setup an upscaler, but i need a controlnet model. I clonned this repository "https://huggingface.co/lllyasviel/ControlNet-v1-1/tree/main" into "models/controlnet" and restarted comfyUI, but it cannot find the model, am I oding something wrong?

[heady grove]

I have downloaded this file from git hub ,

And then extract to my new folder named Forge-ui

[maiden coral]

Isn't the torch same for everything? torch-2.2.2-cp310-cp310-win_amd64.whl

at least it works on A111 (which crashes, so I don't use it anymore)

What are the requirements for A111 forge?

[heady grove]

The end of the error cmd

[heady grove]

Eh how to check this

I have no idea what is torch

[maiden coral]

do everything anew

pip unistall -r requirements.txt or whatever you have

and install them again, check if there are versions e.g. torch ==2.2.2 etc.

it is Machine learning framework which causes troubles all the time

[heady grove]

That's mean I need to delete and redownload ya?

[maiden coral]

well, you can or just make a script for download and install

I generated mine through chatgpt

[heady grove]

Ohh okay

[maiden coral]

I can send them if you want them

[heady grove]

I prefer more easy way hahah

[maiden coral]

but you need to put the .bat files into folder with "requirements.txt" or edit the script

well, I had to do it multiple times, so instead of downloading 3 GB of data I made a script

[heady grove]

For download forge

Click this right ?

[maiden coral]

what to click

just uninstall everything except the python 3.10.x version

and download the repository, if it cotntains requirements txt file then install it

it doesn't take that much time if you have downloaded all the packages into some folder

[vocal burrow]

incorrect command line argument

dunno what you have in your webui-user.bat but it s broken

[heady grove]

Now my brain is error

Now I reinstall everything

[maiden coral]

that's why do everything over again

an just add simple arguments such as --low-vram

if you don't have those 12GB+ video cards

[heady grove]

I add --medvram in a1111 arg

[maiden coral]

and did it work with medvram? What is your GPU?

because I always got CUDA error until I switched to low-vram

[heady grove]

Yes it work , rtx4070 laptop 8gvram

[maiden coral]

ok, then try it

[vocal burrow]

--medvram-sdxl instead

add --cuda-malloc --cuda-stream --pin-shared-memory for extra perfs

[maiden coral]

What does cuda malloc do?

[vocal burrow]

cf https://github.com/lllyasviel/stable-diffusion-webui-forge?tab=readme-ov-file#forge-backend

[maiden coral]

I know what I tried it and it gave me nothing but CUDA error

then just set a basic tags, focus on speed and performance later

[heady grove]

My a1111 bat file

[vocal burrow]

don t put xformers if you re using forge

[heady grove]

Hmmm

So I just cancel it and run the run file again ?

[vocal burrow]

sure

[uncut jewel]

??

[heady grove]

Btw is the
set Phyton = "•••
That line is a must so far I can run my a1111 without it

[vocal burrow]

probably because you have multiple python or you didn t check the "add to PATH'' when installing it

[vocal burrow]

Not enough information, croped error message is too generic.

[heady grove]

Hrmmm

Okay now I just delete that line behind python and --xformers

[uncut jewel]

In order to use SD3, we need to have a stability key ?

[vocal burrow]

for now yes SD3 is only accessible through API

[static mural]

A handsome girl, wearing a black leather jacket, stands in the ruins of a desolate cyberpunk city, in the ruined city, back view, cyberpunk style, multi-color combination ¯_(ツ)_/¯

[verbal burrow]

It doesn't show up in my Lora tab for for the interface

Aye, nevermind

CivitAI was using my filters from the site and wouldn't display because of it

[vocal burrow]

#1047610792226340935

[thorny mortar]

I'm training a new art style image model with 320 sample images. What hyperparameters should I use for training, including epochs, learning rate, text encoder learning rate, and learning rate scheduler?

[karmic crown]

You are using an external drive, so check out this:

https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/13391

[twin wharf]

hi guys! im using zluda for an amd gpu on sd.next, and so far it has been working well but recently my generation speed has been like 10 times slower. sometimes when i restart my pc it works better but then after a while it gets slower again. anyone know what to do?

[ornate elk]

Hey, have you done the additional caching steps of SD.next from their guide?
When this cache gets full it will be slow

[twin wharf]

never heard of this. where would i find the info?

thanks for responding btw

[twin wharf]

i cant find it on the github

[sinful needle]

do you have any ideas what could be causing this? I did some research and a huge number of people are also having the same issue, but there seems to be no one providing any solutions

[ornate elk]

Okay then you didn't enabled it.
What's your GPU BTW?

[twin wharf]

RADEON RX 6950XT

[twin wharf]

how would i enable it?

[ornate elk]

You shouldn't enabled it. As it causes issues. Thats why I asked

[twin wharf]

ohhhh thanks

how do i get rid of the cache making it slow? is that even possible

[ornate elk]

Do you use sdxl models?

[twin wharf]

yes

i never really use sd 1.5

[ornate elk]

Okay

[sinful needle]

Has anyone ever had the error for animatediff?:

*** Error calling: C:\Users\place\Desktop\Studio\Visual\Image\Ai\Stable Diffusion\stable-diffusion-webui\extensions\sd-webui-animatediff\scripts\animatediff.py/ui
Traceback (most recent call last):
File "C:\Users\place\Desktop\Studio\Visual\Image\Ai\Stable Diffusion\stable-diffusion-webui\modules\scripts.py", line 528, in wrap_call
return func(*args, **kwargs)
File "C:\Users\place\Desktop\Studio\Visual\Image\Ai\Stable Diffusion\stable-diffusion-webui\extensions\sd-webui-animatediff\scripts\animatediff.py", line 43, in ui
from scripts.animatediff_mm import mm_animatediff as motion_module
ModuleNotFoundError: No module named 'scripts.animatediff_mm'

[ornate elk]

I can't help much with SD.next as it does so many things different than Auto1111.
If you want you can use my Zluda install guide for Automatic1111.
And then you can test if its slowing down there too.
The guide is in the pinned messages of this channel.
As you already have SD.next working you can skip the hip SDK and zluda folder part

[twin wharf]

ok ill check it out, ty!

[ornate elk]

I'm on AMD too and only use the zluda auto1111 as it has better extension support and easier UI ^^
Haven't had any slowdown problems in the long run

[trail aspen]

Can anyone help me .. I am getting this when trying to install stable iffusion ERROR: Could not find a version that satisfies the requirement torch==2.1.2 (from versions: none)
ERROR: No matching distribution found for torch==2.1.2

[karmic crown]

you are using pyton 3.12

you need 3.10

[trail aspen]

I have 3.10 ... but I downloaded it after downloading 3.12

and i have a feeling it messed something up

[karmic crown]

remove venv folder

and run a1111 again

it will create new venv for you

[trail aspen]

venv ?

[karmic crown]

venv folder

inside your stable-diffusion-webui folder

[trail aspen]

ok

Thank You

did not work

removed everything and restarted ... still getting the same thing

[past willow]

Bump

[maiden coral]

just install it manually

[trail aspen]

How do i do that

[maiden coral]

What idd ozu use to install i?

I generated a script for downloading these files from requirements.txt

through chatgpt

[daring jasper]

Hi, everyone) I'm using SD automatic not that long and not strong enough in coding) so i need help, pleeeese ) desturbing moments are marked in pink

[trail aspen]

yeah I am not very tech savy

I am just getting into this

I have no clue

[maiden coral]

chatgpt can generate a script for downloading and installing files and it works quite well

althoug it had some errors in it

DId you use: pip install torch?

just install the correct version form requirements.txt

[sinful needle]

lets gooooooooooooooooo

[maiden coral]

if it doesn't do it just do pip download requirements into a folder

and then pip install form that folder, but htat's what the scripts are for

you must use version 3.10 of Python

[viral knot]

python 3.11 has a lot of issues, stay with 3.10 as max for now

[sinful needle]

animatediff just needed a new updated commit apparently. their feed was spammed with the same error i had by other users

[ornate elk]

Hey update your extensions and then restart the webui

[sinful needle]

noticing a different error thats unrelated and not affecting too much but worth noting:

C:\Users\place\Desktop\Studio\Visual\Image\Ai\Stable Diffusion\stable-diffusion-webui\extensions\stable-diffusion-webui-two-shot\scripts\two_shot.py:397: GradioDeprecationWarning: The style method is deprecated. Please set these arguments in the constructor instead.

different extension

[vocal burrow]

You can ignore this one

[sinful needle]

what does that mean

the erorr

curious

[viral knot]

maybe old version of gradio

[sinful needle]

but can ignore. gotcha

thanks

[viral knot]

yea

[karmic crown]

2.1.2 does not exists for python 3.12

[maiden coral]

downgrade python first of all

[viral knot]

i wouldnt recommend going past 3.10

[daring jasper]

i'll try! thanks!)

[maiden coral]

to version 3.10 and during the installation process click "ADD TO PATH"

[sinful needle]

Got another one:

*** Error loading script: main.py
Traceback (most recent call last):
File "C:\Users\place\Desktop\Studio\Visual\Image\Ai\Stable Diffusion\stable-diffusion-webui\modules\scripts.py", line 508, in load_scripts
script_module = script_loading.load_module(scriptfile.path)
File "C:\Users\place\Desktop\Studio\Visual\Image\Ai\Stable Diffusion\stable-diffusion-webui\modules\script_loading.py", line 13, in load_module

script_loading.py?

[karmic crown]

remove all python installations you have first
than download and install 3.10 (64bit)
and do not forget to use add to path option

then remove venv folder again, and run webui

[viral knot]

this is why people switch from a1111 🙂

[sinful needle]

wat to comfy ui?

[viral knot]

literally anything else

[sinful needle]

wats the best one because i am not using comfy. i dont know how to logic gate. not wanting a crash coarse on that much extra right now just trying to make nice imagery

[viral knot]

best is subjective, i personally recommend comfy for 3 main reasons: memory efficient, has the most plugins compared to others, has the latest tech sooner than others

[maiden coral]

it'll become easy, I'm 2nd day in, it can even work on 4 GB VRAM, but upscaling is a bit tricky

[karmic crown]

i would not recommend comfy to people who never worked with node-based interfaces, nor to people who are new to SD

[viral knot]

i mean that is fair

[sinful needle]

yeah nodes are what kills me. im kinda smooth brain when it comes to that

[maiden coral]

A111 has problems with crashing, I haven't solved the problem and nothing on the internet works

got "Conentioned errored out" error message

[karmic crown]

you can try forge

it is modified version of a1111

[sinful needle]

less errors?

[viral knot]

all i know is that @maiden coral learned comfy quickly 🙂

[maiden coral]

I wouldn't say, there may still be errors, comfy jsut gives you more freedom.

[viral knot]

yea comfy has more freedom and combinations you can do

[maiden coral]

there are tutorials on the internet for comfy, quite a lot of them, especially for fingers restorationetc

[sinful needle]

well im a graphic design major who switched from web design lets put it that way. im a brainlet

[viral knot]

sad cat

but maybe one day eh :3

[maiden coral]

forge might work too, how RAM efficient is that?

[sinful needle]

ill look into forge

[maiden coral]

VRAM and RAM efficient*

[karmic crown]

@viral knot btw. with default parameters a1111 works faster than both forge and comfy on my m3 pro mac 🙈

[viral knot]

ok now...

lol

[sinful needle]

Forge it is! I'll report back about errors and whether if it fixes them or not. gotta go back up my Loras now

[maiden coral]

are you sure about that, it ate a lot of resources on my notebook

and started crashing

[karmic crown]

after some tweaking forge finally works faster than a1111, but i haven't tried to find better options for comfy yet. didn't had time to play a lot with sd recently

[sinful needle]

im on a future proofed rig so that wont bother me. thats my only saving grace

[maiden coral]

btw how do people include 2 or more lora's in Comfy?

[viral knot]

you chain them

there are also nodes that have multiple lora in one

[maiden coral]

there's a node for lora's, right?

it isn't like a prompt lora with trigger words, is it?

[viral knot]

well not all loras need triggers

[karmic crown]

if you know how to use a1111, switching to comfy should be relativly easy as soon as you understand how flow goes

[viral knot]

yea

most loras you dont even need the clip part, just use the model (unet) connector

[karmic crown]

samplers - same
steps - same
loras - same
...
just the way how you use them is diffrent

i need to go now

cu later

[maiden coral]

bye

[viral knot]

@maiden coral oh i just checked, scott indeed has lcm covered, so check it out maybe https://www.youtube.com/watch?v=pw8eNpmhT5o

it uses a lora too, so you can kill two birds with one stone 🙂

[maiden coral]

whoa, I'll check it out

[waxen pebble]

im trying to figure out how to do git captioning on Kohya ss

getting this error tho

raise ValueError(
ValueError: Unrecognized configuration class <class 'transformers.models.vision_encoder_decoder.configuration_vision_encoder_decoder.VisionEncoderDecoderConfig'> for this kind of AutoModel: AutoModelForCausalLM.

im using the model

https://huggingface.co/nlpconnect/vit-gpt2-image-captioning

btw

blip is working fine... but ive read that git often can do more spesfic captions

[vocal burrow]

it means "'two_shot" is a very old extension that has not been updated in months and is using outdated/deprecated gradio functionalities.

so in the long term it's gonna break at some point whenever gradio decide to remove whatever deprecated function it's using

[sinful needle]

well i dont have to worry about that anymore since im now using forge. ill see how that goes. tired of playing catchup just to normally use the thing

[maiden coral]

Even with 10 steps in SD upscaler It still does more than 10 steps, any ideas? In KSampler I have 11, fixed control after generate and

• Ultimate SD Upscaler has 10 steps now instead of 25, it still goes above 10.

[obsidian carbon]

Anyone good with Lora training here?

[sinful needle]

HOLY SHIT controlnet is integrated into forge? SUPERIOR

FUCK! animatediff is not working still. got a more simple error now: TypeError: 'NoneType' object is not iterable.

the tab is there now which is great, i just can't run it without that error showing up.

[fallen quest]

As far as I remember it tiles your image in smaller images. so it generates (10 Steps x number of tiles)

[maiden coral]

thanks, I was just wondering why so many steps

[sinful needle]

Here's the full error I'm getting with animatediff on Forge:

https://pastebin.com/0uqLf3ZW

It's reading that I don't have the motion module, but it's sitting right there in the folder.

[viral knot]

does Forge have their own discord? maybe you can ask for help there idk :3

[karmic crown]

we have a forge channel on deforum server

but i thing it would be best to raise a ticket on animediff github

[sinful needle]

bro

[viral knot]

time to switch to comfy 🙂

[ornate elk]

From forge webui on github:

[sinful needle]

oh so its actually down

[ornate elk]

Forge need this extension:
https://github.com/continue-revolution/sd-forge-animatediff

You can't use the normal AnimateDiff extension

[sinful needle]

it said under construction so i assume it will stil lbe down regardless right

so that means i cant use the "available" button and just download extensions in the UI of forge? I have to manually grab it somewhere right?

[ornate elk]

I think it should work, as this "under construction" was there since release of forge

[ornate elk]

You have to install it via the extensions tab by using "install from URL"

[sinful needle]

Understood. Giving it a go. Will report back. Thanks.

[ornate elk]

But before that delete the other AnimateDiff

[sinful needle]

yep

keeping the motion module tho as thats crucial

[sinful needle]

noticed after start when i removed Animatediff, the startup text reads "cuda stream activate: false". ignore?

[ornate elk]

Idk, I don't use forge

[sinful needle]

oh. comfy?

[ornate elk]

Auto1111 with AMD

But I can test with my old 1080 later

[sinful needle]

are there advantages with sticking with the ol automatic1111

[ornate elk]

Full extension support would be one.
Regular Updates and more Developers behind it.
The Dev of Forge is also the dev of controlnet and Fooocus webui so he got to much to work on. He didn't updated anything in the last 1-2 months. But he works on a hidden project.
(That shouldn't be negativ on him, he just has to much to work on ^^)
And auto1111 is better compatible with AMD.

[sinful needle]

i think the lora settings are the same in both forge and auto. i forget how to make it so the lora name in the text prompt is the same as what ive named the file

also animatediff works now thanks

[ornate elk]

Nice 👍

[ornate elk]

Yes they should be the same.
In auto1111 it was under settings

[merry scroll]

Hey everyone,

Any ideas for SDXL Turbo to produce non-double heads?

[karmic crown]

use some model from civitai that is based on turbo instead of original one

and use recommended dimensions

[merry scroll]

the official release is broken?

its from stability ai

[karmic crown]

no, but custom models are better

[merry scroll]

okay, will have a look

[karmic crown]

https://civitai.com/models/112902/dreamshaper-xl?modelVersionId=351306

for example

[merry scroll]

any recommendation maybe? usually need photorealistic or cinematic stuff

oh thanks

[karmic crown]

set filter like this and take what you like

i personally like lightning much more than turbo

[merry scroll]

is lightning faster than turbo? (sounds like it 😄 )

[karmic crown]

it create great images in 2, 4 or 8 steps

i use 8 version

https://huggingface.co/ByteDance/SDXL-Lightning

this is original version

i use this one the most

https://civitai.com/models/202108/bluepencil-xl-lcm-lightning?modelVersionId=369896

[merry scroll]

turbo works with 1 step too so its quite fast

except the double head issue 😄

[karmic crown]

lightning is fast as well, but creates great images

[merry scroll]

will try both and keep the one that feels better

[karmic crown]

8 steps

this one is 4 steps only

https://civitai.com/models/112902/dreamshaper-xl?modelVersionId=354657

[sonic pine]

its not external

[karmic crown]

hmm... i only found that with same error you had

[sonic pine]

its not that bad ig

pictures still show up

just not on the web ui

[bronze spade]

Hello friend, i can't see the pix2pix on my controlNet , i dowloaded using url github but nothing show , can someone help me ? thanks

i've Forge UI

[fair oxide]

my a1111 wont launch, it just pulls this up and doesnt do anything else

[stark mist]

Hi... New to the server... I have an issue with my SD (Automatic1111). After I updated (via the extensions tab in SD) all the ControlNet options are just gone.... the startup cmd field shows plenty other errors as well.... Am I right here, or should i change to another part of the server?

[fair oxide]

nvm, i fixed it lool

[hardy canopy]

Any idea why A1111 has started stalling out around 96% and never recovers?
I can usually get 1-2 generations right off the bat after a restart, then this happens
My whole PC grinds to a halt when this happens. I've tried waiting it out and sometimes it takes 10-15 min. Sometimes it times out.

[sinful needle]

should i git clone https://github.com/deforum-art/sd-forge-deforum Deforum directly into the forg webui folder or somewhere else?

[zealous ore]

After installing Facechain I get "ERROR facechain: failed to install mmcv, make sure to have "CUDA Toolkit" and "Build Tools for Visual Studio" installed. I have CUDA 11.7 installed. Used to have 11.8 same error. I installed the CUDNN files in the coorosponding folder in Program Files. I installed Visual Studio Build Tools 2022 along with the Desktop development with C++, MSVC v143 VS 2022 C++ Build Tools, Windows 11 SDK, C++ CMake Tools and C++/CLI support for v143 build tools.

[ornate elk]

hey, please follow my install guide from the pinned messages of this channel.
then it should work
you installed an outdated version inside the wrong folder

[ornate elk]

whats your GPU?
Whats inside your wbeui-user.bat?
how much RAM?

[ornate elk]

into the extensions folder

[maiden coral]

Has anyone tried ReActor Node in ComfyUI? That node doesn't to work, probably something up with a branch.

It is for swapping faces.

[heady grove]

Can forge UI share the extension with a1111 ? Just like other models

[subtle quarry]

none of the preprocessors are working. urllib.error.URLError: <urlopen error [WinError 10054] An existing connection was forcibly closed by the remote host>
Using forge

google isn't helping, any help would be nice.

[ornate elk]

not every extension that works on auto1111 works on forge,
but you can copy the extensions from the extensions folder of auto1111 to forges

[ornate elk]

thats the wrong openpose model

[subtle quarry]

model works fine, preprocessor does nothing

[heady grove]

Oh okay

Tq

[ornate elk]

okay, the connection error can come from your Antivirus programm or firewall, proxy, VPN, or DNS

[subtle quarry]

..awesome. i love when the firewall decides to be a jerk

[ornate elk]

its mostly not the windows firewall that blocks

[subtle quarry]

it doesn't let me access hugging face 90% of the time

[ornate elk]

then change your DNS server

[subtle quarry]

how do I do that?

...google

didn't do anything

unless i have to restart the computer

[ornate elk]

you have to reset your dns entry by open up a cmd and type
ipconfig /flushdns

then restart the browser

[subtle quarry]

ahhh.

yeah that did jack nothing

i'll still give the other thing a shot.

yep. same error

urllib.error.URLError: <urlopen error [WinError 10054] An existing connection was forcibly closed by the remote host>

[heady grove]

Where is the preset file

Location

[subtle quarry]

me?

[heady grove]

No

I'm just asking hahaha

[ornate elk]

do you have a third party antivirus or adblocker?
make sure the webui is whitelisted in any adblocker

[subtle quarry]

ah. adblocker, i use uh... opera gx so that might be the issue

[heady grove]

[subtle quarry]

error again

[ornate elk]

yea try an other browser, and make sure OperaGX VPN is disabled

[subtle quarry]

error in firefox

this is fucking annoying, i dont know what's wrong with it

[noble belfry]

is this to get a model from HF?

[subtle quarry]

i have ublock origin, and malwarebytes, but neither have ever been an issue.

[noble belfry]

if so, and this is caused by controlnet trying to download them from the cmd, check in browser if theyre set to require login

[subtle quarry]

no, this is to get a preprocessor working

none of them are working

[noble belfry]

in the cmd window. does it look like its trying to download the files from the host is what i mean

[ornate elk]

What's the error ?

[subtle quarry]

Traceback (most recent call last):
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\gradio\routes.py", line 488, in run_predict
    output = await app.get_blocks().process_api(
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\gradio\blocks.py", line 1431, in process_api
    result = await self.call_function(
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\gradio\blocks.py", line 1103, in call_function
    prediction = await anyio.to_thread.run_sync(
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\anyio\to_thread.py", line 33, in run_sync
    return await get_asynclib().run_sync_in_worker_thread(
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\anyio\_backends\_asyncio.py", line 877, in run_sync_in_worker_thread
    return await future
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\anyio\_backends\_asyncio.py", line 807, in run
    result = context.run(func, *args)
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\gradio\utils.py", line 707, in wrapper
    response = f(*args, **kwargs)
  File "C:\Users\Quinn\Desktop\Forge\webui\extensions-builtin\sd_forge_controlnet\lib_controlnet\controlnet_ui\controlnet_ui_group.py", line 847, in run_annotator
    result = preprocessor(
  File "C:\Users\Quinn\Desktop\Forge\webui\extensions-builtin\forge_legacy_preprocessors\scripts\legacy_preprocessors.py", line 103, in __call__
    result, is_image = self.call_function(img=input_image, res=resolution, thr_a=slider_1, thr_b=slider_2, **kwargs)
  File "C:\Users\Quinn\Desktop\Forge\webui\extensions-builtin\forge_legacy_preprocessors\legacy_preprocessors\preprocessor.py", line 291, in run_model
    return remove_pad(self.model_openpose(
  File "C:\Users\Quinn\Desktop\Forge\webui\extensions-builtin\forge_legacy_preprocessors\annotator\openpose\__init__.py", line 447, in __call__
    poses = self.detect_poses(oriImg, include_hand, include_face)
  File "C:\Users\Quinn\Desktop\Forge\webui\extensions-builtin\forge_legacy_preprocessors\annotator\openpose\__init__.py", line 333, in detect_poses
    self.load_model()
  File "C:\Users\Quinn\Desktop\Forge\webui\extensions-builtin\forge_legacy_preprocessors\annotator\openpose\__init__.py", line 208, in load_model
    load_file_from_url(body_model_path, model_dir=self.model_dir)
  File "C:\Users\Quinn\Desktop\Forge\webui\modules\modelloader.py", line 39, in load_file_from_url
    download_url_to_file(url, cached_file, progress=progress)
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\torch\hub.py", line 620, in download_url_to_file
    u = urlopen(req)
  File "C:\Program Files\Python310\lib\urllib\request.py", line 216, in urlopen
    return opener.open(url, data, timeout)
  File "C:\Program Files\Python310\lib\urllib\request.py", line 519, in open
    response = self._open(req, data)
  File "C:\Program Files\Python310\lib\urllib\request.py", line 536, in _open
    result = self._call_chain(self.handle_open, protocol, protocol +
  File "C:\Program Files\Python310\lib\urllib\request.py", line 496, in _call_chain
    result = func(*args)
  File "C:\Program Files\Python310\lib\urllib\request.py", line 1391, in https_open
    return self.do_open(http.client.HTTPSConnection, req,
  File "C:\Program Files\Python310\lib\urllib\request.py", line 1351, in do_open
    raise URLError(err)
urllib.error.URLError: <urlopen error [WinError 10054] An existing connection was forcibly closed by the remote host>```

[ornate elk]

What's your DNS setting?

[subtle quarry]

i changed it back, but it was 8.8.8.8

[ornate elk]

Okay and you whitelisted the webui in ublock?

[subtle quarry]

i doubt that's the issue, it's worked just fine.

but i'll give it a shot i guess.

i'll change the dns again, the error is the same.

[ornate elk]

If your from asia, Russia, Iran you may need to use a VPN

[subtle quarry]

America.

[ornate elk]

Okay

[subtle quarry]

error is the same :/

[ornate elk]

Then maybe restart the PC too.
Also you didn't said what's your antivirus programm

[subtle quarry]

malwarebytes

but it's never been an issue.

[ornate elk]

Okay

Whitelist the webui in malware bytes browser add-on too and maybe try disable the realtime protection for test

[noble belfry]

i mean the http req is coming from python, its not coming from or through the browser. and seeing half of it in program files and the other half on desktop is making my teeth itch too*. thats half user folders half system

[subtle quarry]

well! i'm sorry im not perfect in the file organization department.

[noble belfry]

oh its not that, its that windows is whiny as hell about permissions

but silently. itll just throw errors and you never know if thats a culprit

[subtle quarry]

don't know how to do either of those but i'll try just turning off the firewall entirely because why not.

yay! same error

what the hell is thing's problem

everything else works FINE just not the preprocessors.

[rocky adder]

Help please!
In the middle of a generation with Automatic1111 my computer got turned off
Now it just says "Error" in the Stable Diffusion Checkpoint field, and it will simply not work
After Restarting computer, now Automatic1111 doesn't even launch
What can I do to correct this? I guess turning off my computer in the middle of generating images did mess up Automatic1111

[ornate elk]

Hey, what's your GPU?
And whats inside your webui-user.bat?

[rocky adder]

Nvidia 3090
@echo off

set PYTHON=
set GIT=
set VENV_DIR=
git pull
COMMANDLINE_ARGS=

call webui.bat

[ornate elk]

Okay at the line
Commandline_args=
Add: --xformers
Then delete the venv folder and relaunch the webui-user.bat

[noble belfry]

you can download those manually. but it would be NICE to do it programatically

[subtle quarry]

the preprocessors ARE downloaded.

they just wont DO anything.

[ornate elk]

Do you have automatic 1111 still installed?

[subtle quarry]

it's on my partner's side of the computer, but yes

[ornate elk]

Ah okay

[subtle quarry]

i can still give the preprocessor a shot on their side though

[ornate elk]

Because I often heard of Forge users with Controlnet errors.
And the best way was to just let it uses the models from auto1111

make sure you use these controlonet models for 1.5:
https://civitai.com/models/38784/controlnet-11-models

[subtle quarry]

it's the same error

so there's something else going on here or whatever

also i'm using sdxl

so

[ornate elk]

ohhh, do you have the right sdxl controlnet models?

you can find them here:
https://huggingface.co/lllyasviel/sd_control_collection/tree/main

[subtle quarry]

Those are massive

and again, hugging face is an ass to me

[ornate elk]

those are the ones that work with sdxl

[subtle quarry]

I believe it was this one https://huggingface.co/OzzyGT/controlnet-openpose-sdxl-1.0

i dont know how to remove embed

[ornate elk]

try out that one:

[subtle quarry]

again, i dont think it's the model. its the preprocessor.

the model works fine.

there's no issue with it.

[ornate elk]

how do you know if it doesnt work?

or did you used it in auto1111?

[subtle quarry]

I can use the model.

I can't use the preprocessor.

The preprocessor is throwing the error.

WHen I use controlnet normally, it works.

[noble belfry]

that error is a URLLIB request to download a file that is not working.

[subtle quarry]

when i try to run an image through the preprocessor, it throws an error.

[ornate elk]

yea because something on your system is blocking access to hugginface where the preprocessor file is stored

[subtle quarry]

i don't know what is though!

I've turned off the firewall, changed DNS

[rocky adder]

Wow, thanks lot for such quick response... I thought it would take days...
I just did everything you told me, unfortunately it just doesn't launch
It says:

[ornate elk]

hey, thats a connection error. something could block the webui. make sure to whitelist the webui in any browser adblocker

[ornate elk]

have you restarted the PC at least once?

[subtle quarry]

like 3 times

[ornate elk]

can you access huggingface in your browser?

[subtle quarry]

90% of the time, no.

[ornate elk]

try with firefox

and send the error you get

[subtle quarry]

Traceback (most recent call last):
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\gradio\routes.py", line 488, in run_predict
    output = await app.get_blocks().process_api(
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\gradio\blocks.py", line 1431, in process_api
    result = await self.call_function(
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\gradio\blocks.py", line 1103, in call_function
    prediction = await anyio.to_thread.run_sync(
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\anyio\to_thread.py", line 33, in run_sync
    return await get_asynclib().run_sync_in_worker_thread(
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\anyio\_backends\_asyncio.py", line 877, in run_sync_in_worker_thread
    return await future
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\anyio\_backends\_asyncio.py", line 807, in run
    result = context.run(func, *args)
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\gradio\utils.py", line 707, in wrapper
    response = f(*args, **kwargs)
  File "C:\Users\Quinn\Desktop\Forge\webui\extensions-builtin\sd_forge_controlnet\lib_controlnet\controlnet_ui\controlnet_ui_group.py", line 847, in run_annotator
    result = preprocessor(
  File "C:\Users\Quinn\Desktop\Forge\webui\extensions-builtin\forge_legacy_preprocessors\scripts\legacy_preprocessors.py", line 103, in __call__
    result, is_image = self.call_function(img=input_image, res=resolution, thr_a=slider_1, thr_b=slider_2, **kwargs)
  File "C:\Users\Quinn\Desktop\Forge\webui\extensions-builtin\forge_legacy_preprocessors\legacy_preprocessors\preprocessor.py", line 291, in run_model
    return remove_pad(self.model_openpose(
  File "C:\Users\Quinn\Desktop\Forge\webui\extensions-builtin\forge_legacy_preprocessors\annotator\openpose\__init__.py", line 447, in __call__
    poses = self.detect_poses(oriImg, include_hand, include_face)
  File "C:\Users\Quinn\Desktop\Forge\webui\extensions-builtin\forge_legacy_preprocessors\annotator\openpose\__init__.py", line 333, in detect_poses
    self.load_model()
  File "C:\Users\Quinn\Desktop\Forge\webui\extensions-builtin\forge_legacy_preprocessors\annotator\openpose\__init__.py", line 208, in load_model
    load_file_from_url(body_model_path, model_dir=self.model_dir)
  File "C:\Users\Quinn\Desktop\Forge\webui\modules\modelloader.py", line 39, in load_file_from_url
    download_url_to_file(url, cached_file, progress=progress)
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\torch\hub.py", line 620, in download_url_to_file
    u = urlopen(req)
  File "C:\Program Files\Python310\lib\urllib\request.py", line 216, in urlopen
    return opener.open(url, data, timeout)
  File "C:\Program Files\Python310\lib\urllib\request.py", line 519, in open
    response = self._open(req, data)
  File "C:\Program Files\Python310\lib\urllib\request.py", line 536, in _open
    result = self._call_chain(self.handle_open, protocol, protocol +
  File "C:\Program Files\Python310\lib\urllib\request.py", line 496, in _call_chain
    result = func(*args)
  File "C:\Program Files\Python310\lib\urllib\request.py", line 1391, in https_open
    return self.do_open(http.client.HTTPSConnection, req,
  File "C:\Program Files\Python310\lib\urllib\request.py", line 1351, in do_open
    raise URLError(err)
urllib.error.URLError: <urlopen error [WinError 10054] An existing connection was forcibly closed by the remote host>```

had to change default browser to firefox to see if that would do anything...clearly not.

[ornate elk]

i said, open the website (huggingface) in firefox

to test if your able to access it

[subtle quarry]

ah

yeah i can

okay now why is it letting me access the site on opera

[noble belfry]

just out of curiousity. try and open forge in administrator

[subtle quarry]

aight

[ornate elk]

yea its probably a big issue that you have python installed in programm files and forge on desktop

[subtle quarry]

running as admin opens the cmd for a split second then crashes it

:/

[rocky adder]

Thanks! But I have changed nothing in my settings! Also I don't have any addblocker of any type, I am now in firewall settings, and can't add webui neither since it's not an app.
Now I changed my deffault browser to a new one, and its the same:
As I said, Automatic 1111 just got broken when my computer turned off in the middle of a Deforum image generation process

[subtle quarry]

and hugging face isn't working again

[noble belfry]

crashes or closes? right click on the bat file

[subtle quarry]

it closes, sorry

hugging face isn't working again anyway

so i think i need to figure out why

[ornate elk]

okay then try the following:
go into the extensions folder and delete:
animator_extension
and Deforum
then delete the venv folder again and relaunch the webui-user.bat

[subtle quarry]

i moved forge to program files and it did jack nothing

this is really getting on my fucking nerves

[stuck knot]

Hi, I'm facing an issue with a ComfyUI update where text strings enclosed in curly brackets {} are no longer working at all. This problem disrupts my output text and images by producing incorrect results, does anyone know how to fix it?

[rocky adder]

Thanks a lot for your super quick responses ... I just did everything you said with same results:

[noble belfry]

oh jesus dont put it in program files. those are completely different permissions

[subtle quarry]

well i moved it out

so

[noble belfry]

you'll have those permissions stuck on the directory now though

[subtle quarry]

it works fine

idfk know man

im getting pissed off and I Just want this function to work

this is the error i get with hugging face

[noble belfry]

intermittant?

[subtle quarry]

i guess.

[noble belfry]

does the like of say, youtube only load half the thumnails for things or anything else like that?

[subtle quarry]

it works, it doesn't, it works again, it doesn't

youtube works fine

every other website works just fine

it's legit just hugging face that's been giving me issues.

[noble belfry]

what country are you in?

[subtle quarry]

America

No vpn either

[noble belfry]

oh then its definately not them doing it for badness on your conn

[subtle quarry]

I dont know what the issue is, i can try it on my laptop

[ornate elk]

okay, thats really strange.
guess some files are messed up/broken from the crash.
also your on python 3.10.7, you should update that to 3.10.11 64bit
i would suggest making a clean install following my guide.
then you can move your stuff over if it works.
my guide: #🤝｜tech-support message

[subtle quarry]

okay so it works fine on my laptop

both are plugged in via ethernet

[maiden coral]

Have you tried uninstallation? Also why huggin face? What is the exact problem?

what did you change before the it started happening?

[subtle quarry]

huggingface for the controlnet preprocessors. clean install will do nothing, as it's computer wide.

huggingface is just blocked for some damn reason.

i've tried turning off the firewall, changing the dns, setting up custom rules for the firewall

it's done absolutely nothing.

[maiden coral]

https://huggingface.co/lllyasviel/ControlNet this thing?

[subtle quarry]

i feel like im going in circles

[maiden coral]

I only know that huggingface is a website for these things.

[subtle quarry]

The error.

Traceback (most recent call last):
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\gradio\routes.py", line 488, in run_predict
    output = await app.get_blocks().process_api(
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\gradio\blocks.py", line 1431, in process_api
    result = await self.call_function(
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\gradio\blocks.py", line 1103, in call_function
    prediction = await anyio.to_thread.run_sync(
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\anyio\to_thread.py", line 33, in run_sync
    return await get_asynclib().run_sync_in_worker_thread(
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\anyio\_backends\_asyncio.py", line 877, in run_sync_in_worker_thread
    return await future
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\anyio\_backends\_asyncio.py", line 807, in run
    result = context.run(func, *args)
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\gradio\utils.py", line 707, in wrapper
    response = f(*args, **kwargs)
  File "C:\Users\Quinn\Desktop\Forge\webui\extensions-builtin\sd_forge_controlnet\lib_controlnet\controlnet_ui\controlnet_ui_group.py", line 847, in run_annotator
    result = preprocessor(
  File "C:\Users\Quinn\Desktop\Forge\webui\extensions-builtin\forge_legacy_preprocessors\scripts\legacy_preprocessors.py", line 103, in __call__
    result, is_image = self.call_function(img=input_image, res=resolution, thr_a=slider_1, thr_b=slider_2, **kwargs)
  File "C:\Users\Quinn\Desktop\Forge\webui\extensions-builtin\forge_legacy_preprocessors\legacy_preprocessors\preprocessor.py", line 291, in run_model
    return remove_pad(self.model_openpose(
  File "C:\Users\Quinn\Desktop\Forge\webui\extensions-builtin\forge_legacy_preprocessors\annotator\openpose\__init__.py", line 447, in __call__
    poses = self.detect_poses(oriImg, include_hand, include_face)
  File "C:\Users\Quinn\Desktop\Forge\webui\extensions-builtin\forge_legacy_preprocessors\annotator\openpose\__init__.py", line 333, in detect_poses
    self.load_model()
  File "C:\Users\Quinn\Desktop\Forge\webui\extensions-builtin\forge_legacy_preprocessors\annotator\openpose\__init__.py", line 208, in load_model
    load_file_from_url(body_model_path, model_dir=self.model_dir)
  File "C:\Users\Quinn\Desktop\Forge\webui\modules\modelloader.py", line 39, in load_file_from_url
    download_url_to_file(url, cached_file, progress=progress)
  File "C:\Users\Quinn\Desktop\Forge\webui\venv\lib\site-packages\torch\hub.py", line 620, in download_url_to_file
    u = urlopen(req)
  File "C:\Program Files\Python310\lib\urllib\request.py", line 216, in urlopen
    return opener.open(url, data, timeout)
  File "C:\Program Files\Python310\lib\urllib\request.py", line 519, in open
    response = self._open(req, data)
  File "C:\Program Files\Python310\lib\urllib\request.py", line 536, in _open
    result = self._call_chain(self.handle_open, protocol, protocol +
  File "C:\Program Files\Python310\lib\urllib\request.py", line 496, in _call_chain
    result = func(*args)
  File "C:\Program Files\Python310\lib\urllib\request.py", line 1391, in https_open
    return self.do_open(http.client.HTTPSConnection, req,
  File "C:\Program Files\Python310\lib\urllib\request.py", line 1351, in do_open
    raise URLError(err)
urllib.error.URLError: <urlopen error [WinError 10054] An existing connection was forcibly closed by the remote host>```

[maiden coral]

I usually jjust do git clone

You cannot do git clone?

[subtle quarry]

i doubt i can if hugging face is throwing a fit

[maiden coral]

You cannot load the website?

[subtle quarry]

no.

[maiden coral]

Like at all?

[subtle quarry]

90% of the time it's blocked.

[maiden coral]

Can you show me the screenshot?

[maiden coral]

reset the firewall to default state first of all

Have you tried to delete cache?

or use CCleaner which is free

Does the website work when you use data of your phone if you have any or if it works on your phone on the same Wi-Fi connection?

try a different device and see if it works

[noble belfry]

do not use ccleaner. that thing will ruin your registry.

[subtle quarry]

it works on my laptop, on the same internet

ethernet

[maiden coral]

no, it doesn't ruin registry

it's the Windows that is being ruined with every update

[subtle quarry]

im not using ccleaner

[noble belfry]

what about in your partners login?

[subtle quarry]

it doesn't work

it's the entire computer.

something here is blocking it and i can't for the life of me figure out wtf it is

i reset the firewall to default and it did jackshit--aka NOTHING.

Traceback (most recent call last):
  File "C:\Program Files\Python310\lib\urllib\request.py", line 1348, in do_open
    h.request(req.get_method(), req.selector, req.data, headers,
  File "C:\Program Files\Python310\lib\http\client.py", line 1282, in request
    self._send_request(method, url, body, headers, encode_chunked)
  File "C:\Program Files\Python310\lib\http\client.py", line 1328, in _send_request
    self.endheaders(body, encode_chunked=encode_chunked)
  File "C:\Program Files\Python310\lib\http\client.py", line 1277, in endheaders
    self._send_output(message_body, encode_chunked=encode_chunked)
  File "C:\Program Files\Python310\lib\http\client.py", line 1037, in _send_output
    self.send(msg)
  File "C:\Program Files\Python310\lib\http\client.py", line 975, in send
    self.connect()
  File "C:\Program Files\Python310\lib\http\client.py", line 1454, in connect
    self.sock = self._context.wrap_socket(self.sock,
  File "C:\Program Files\Python310\lib\ssl.py", line 513, in wrap_socket
    return self.sslsocket_class._create(
  File "C:\Program Files\Python310\lib\ssl.py", line 1071, in _create
    self.do_handshake()
  File "C:\Program Files\Python310\lib\ssl.py", line 1342, in do_handshake
    self._sslobj.do_handshake()
ConnectionResetError: [WinError 10054] An existing connection was forcibly closed by the remote host```

i dont know why the hell it's doing this

and it's really getting under my skin.

[noble belfry]

starting to sound like HF is doing it out of spite

[subtle quarry]

it's really pissing me off

i'd rather not do the preprocessor on my laptop, and transfer it over to the pc

[maiden coral]

or you can use wireshark

[subtle quarry]

dudes. if i install any other 3rd party software on this thing my s/o might kill me.

[maiden coral]

or just clean up the PC

I usually clean up my PC after half a year or once a year

because the programs themselves leave fiels in PC and nothing can be done about it

[subtle quarry]

.

[noble belfry]

one last thing to try. release and renew your ip i guess.

[subtle quarry]

can't do that either, s/o will have my neck if this messes up anything.

[noble belfry]

just on the crazy offchance that it really is on the host end

oh then maybe they've put restrictions on it 🙂

[subtle quarry]

i have admin privs

they are not the most tech savvy

[maiden coral]

or maybe your internet provider blocks that IP address

if other websites work fine

[subtle quarry]

nope, because it works fine on my laptop

[maiden coral]

ik

that's why I crossed it out

[noble belfry]

well all release/renew does is get a new ip from your router. but it sounds like suck it up and put it on the laptop.

[maiden coral]

Have you tried AI copilot?

or chatGPT, AI Copilot might be better for it as it has internet access

[noble belfry]

none of this is relevant buddy. you're just saying things.

[maiden coral]

just insert this error message

[subtle quarry]

it's just gonna throw me what google has told me

which is? nothing.

bugs. just people complaining about it and no one having a solution.

so there's jack shit i can do and don't tell me to 'suck it up'

[rocky adder]

Ok! Just two questions before messing around with my system,
I use Unreal Engine 5.1 to pack games, does the installation of Python 3.10.11 interfere somehow with my system?
Also, do I need to install again Git 64bit?

Or should I simply start from point number 3:
Then you copy and paste this command:
git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git

One thousand thanks!

[maiden coral]

well, it gave me some options

definitely not the same as google

[ornate elk]

if you already have git installed then you dont need it.
python shouldnt to anything to your unreal engine 5.1
3.10.11 ist just a version with more bugs fixed so nothing will change. you can just overwrite your 3.10.7 install (upgrade) if asked

[rocky adder]

Thank youuuuuuuu

[maiden coral]

Have you cleared the browser's cache?

[subtle quarry]

for the 5th time. it's something computer wide

[maiden coral]

CCleaner makes these things easier.

ok

[subtle quarry]

it's on both profiles on the computer.

it's not tied to the browser

[maiden coral]

at least show the website's URL

[subtle quarry]

https://huggingface.co

[maiden coral]

and wireshark would help you

at least determine what packets go through and which not

[subtle quarry]

for the 3rd time, i am not putting 3rd party software on ths computer.

[maiden coral]

Have you tried to run your PC in safe mode?

If it works then there might be a program or a virus that blocks it.

safe mode with internet connection

[subtle quarry]

yeah so that didnt work

  File "C:\Users\Quinn\Desktop\Forge\webui\launch.py", line 51, in <module>
    main()
  File "C:\Users\Quinn\Desktop\Forge\webui\launch.py", line 39, in main
    prepare_environment()
  File "C:\Users\Quinn\Desktop\Forge\webui\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
Press any key to continue . . .```

[maiden coral]

What are you trying to do now?

[subtle quarry]

wtf do you think im doing? Launching from safe mode like you said!

[maiden coral]

I thought it was only hugging face website you cannot access.

well, because I see a different error message

[subtle quarry]

yeah no shit!

[maiden coral]

you're clearly trying to use torch and it failed to start

raise RuntimeError(
RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check
Press any key to continue . . .

[subtle quarry]

i give up

[noble belfry]

what they are asking i guess is to try huggingface from a browser in safemode

not forge.

[maiden coral]

well, what i understoood from the initial question is why "huggingface.co" website doesn't work on PC.

then I see forge and torch which isn't a browser

if it's related to torch and forge do a clean install of it

that's the thing that helped me

pip unistall eveyrthing you had from requirements file and then install again with their proper versions

[subtle quarry]

no, no it's not a pytorch issue

idfk what the issue is

but i give up.

im going in circles of "do this, no no do this instead, wait a minute get this software and do some cleaning"

[maiden coral]

yeah, that's the thing that always helps 100%

since oyu clean all the corrupted files in your computer

[solemn atlas]

Hi, i have a question, why does all XL models look like this

[subtle quarry]

change your vae to an sdxl one

[subtle quarry]

well if it's any consolation, it works in safe mode

im not putting 3rd party shit on here, end of discussion

[maiden coral]

yeah, so it must be some ind of software, have you scanned your PC on viruses?

[solemn atlas]

thank you so much, it worked now

[subtle quarry]

it's just gonna be a bunch of false positives

and im too tired to deal with that shit

[maiden coral]

just take a break and sometimes do a cleanup of PC, it helps to solve a lot of problems

installation and quick setup doesn't take more than 20 min. if you're on SSD

[subtle quarry]

If you're telling me to install 3rd party software again

i'm going to throw a fit.

[hardy canopy]

Still haven't been able to figure this out. Generations show as 100% completed in the log, but everything stalls around 96% and my computer bogs down and becomes unusably slow. So bizarre.

• Windows 10, GPU: RTX 3080 w/ 10GB VRAM (drivers updated)
• Browsers: Edge & Chrome (up to date)
• Webuser.bat: COMMANDLINE_ARGS= --api --autolaunch --xformers --opt-channelslast --upcast-sampling --opt-split-attention --skip-torch-cuda-test --skip-python-version-check
• Installed & running on an M.2 SSD
• Using Juggernaut_v9 SDXL
• Not using Hirez- fix or Refiner
• Tried a fresh re-install of A1111
• Tried deleting VENV folder
• No errors in the log

[ornate elk]

whats your gpu ?

[hardy canopy]

GPU: RTX 3080 w/ 10GB VRAM (drivers updated)

[ornate elk]

okay then you should remove: --opt-channelslast --upcast-sampling --opt-split-attention --skip-torch-cuda-test --skip-python-version-check
and instead add: --medvram-sdxl --no-half-vae to the other args (xformers, api)

[hardy canopy]

ty. i'll give that a try!

[rocky adder]

IT WORKED!! :))))
So now what folders should i put in this new installation?? Only my outputs and extensions and ssafetensors and so on?

One thousand thanks, Really

[maiden coral]

you just install it, Git is for colling repositories from github and similar websites (such as huggingface where you can download models etc.)

[ornate elk]

your models/stable-diffusion and models/lora folder,
maybe embeddings if you had them
and of course the outputs folder
please install the extensions again via the extensions tab and dont copy them

[heady grove]

Forge UI is not support tile diffusion and vae ?

[maiden coral]

or you might have to download it, ComfyUI is better in these things

you setup a workflow which you can use infinitely

Does anyone know what is the difference between CLIPTextEncodeSDXL in face refining workflow? What is the differnece between a normal one where L and G isn't? ALso is there an SD version of it?

because I wanted to refine a picture using SD model but it'd throw an error since it's for SDXL.

[ornate elk]

forge has that extension included

but idk if its fully supported

[heady grove]

The UI is not same with multi diffusion on a1111

I don't know how to use it

Ok btw I fixed my problem yesterday by changing " to '

Cna I know what the function of no half vae

I put my picture to a1111 and use tile diffusion and came out this

[maiden coral]

add --no-half argument

[heady grove]

Yaya I add it just now

I want to make a 6k picture with tile diffusion

[ornate elk]

only add --no-half-vae not --no-half

[heady grove]

I have add --no-half-vae --xformers and also --medvram-sdxl

[maiden coral]

Is there a difference between these two?

[heady grove]

Any method I can get 6k resolution in forge ui

[maiden coral]

that's quite a big resolution

[heady grove]

[ornate elk]

yes, --no-half loads every model and vae as fp32 instead of fp16.
--no-half-vae only loads vae files as fp32
--no-half will slow down generation time and uses more vram. Some GPUs like the GTX 16xx Series or the AMD RX5xx Series need it because these cards cant load fp16 by default

[heady grove]

Since forge UI is not support tile diffusion I need to move the picture to a1111

[maiden coral]

thank you, so that slowed down my generation times, I have no such problems on ComfyUI, thank you

[maiden coral]

It doesn't?

[noble belfry]

forge should be automatically tiling when upscaling. an issue with a lot of the things forge "doesnt support" is that theyre integrated already and doubling them up throws errors. in an ideal world you shouldnt be needing any arg switches.

[maiden coral]

you can use ComfyUI and do it there, although it takes time to go from 640x640 to 2560x2560 on faster GPU's it'll be quicker

[heady grove]

Hrmmm

Ya I see a multi diffusion integrated

[noble belfry]

it has most things natively. the secret to get it to work is just assume it knows. if you want to make a big image, just try it. tiled etc should kick in

[heady grove]

But how can i set the final image resolution on multi diffusion forge UI