🤝｜tech-support

[ornate elk]

the setup shouldnt be to difficult, just read the info before starting. It explains how it works.

[tall pier]

@ornate elk thanks mate I got it

[ornate elk]

Perfect no problem 🙂

[tall pier]

now i just have to download checkpoints and loras

[ornate elk]

You can get them on Civitai.com

[tall pier]

Currently using tensor to download since i have used that site to gen and is familiar with those

[weak whale]

For some reason my loras were not showing up after I fixed the problem they are not working.If I try to run a promot with just the lora I get an error

[ornate elk]

Scam

[weak whale]

I know

[ornate elk]

Make sure they are compatible with your model

They only show up if they are

[weak whale]

ok let me try other models

[tame cave]

Hey guys i got an issue, i'm trying to redirect my model folder from another local without having to duplicate models, so i tried to follow from earlier messages

but somehow --models-dir is not recognized

[ornate elk]

hey, which webui do you use?

[tame cave]

what do you mean ?

[ornate elk]

which tool do you want to use with --models-dir? auto1111 or forge or reforge ?

[tame cave]

forge

[ornate elk]

Okay then you need to update it

Run the update.bat and then --models-dir should work

Also use " " around the path

[tame cave]

is there a specific place to put the --model-dir ?

[ornate elk]

In the webui-user.bat at the line set COMMANDLINE_ARGS=

[tame cave]

ok thanks and i think i just got the issues, since my loras and checkpoint folders are on my ComfyUI folders, they have differents names than the SDF ones...

[ornate elk]

yea its better and easier to go from the other way, meaning move everything to forge and link it to comfyui

but in forge you can also define every path on its own in the webui-user.bat

or you can use --ckpt-dir "path to models"
--lora-dir "path to loras"
--controlnet-dir
--embeddings-dir
--vae-dir

[tame cave]

i just done that

and it worked

[ornate elk]

ah good!

[tame cave]

should i be concerned with this message E:\webui_forge_cu121_torch231\system\python\lib\site-packages\transformers\utils\hub.py:128: FutureWarning: Using TRANSFORMERS_CACHE is deprecated and will be removed in v5 of Transformers. Use HF_HOME instead.

[ornate elk]

Nope its normal and can be ignored

[tame cave]

all right thanks then

[ornate elk]

Np

[fiery forge]

is there any way to speed up the image generation? im new to this local stuff and im pretty sure my pc is more than good enough to do this but when i use stable diffusion my computer runs hellllllla slow and the image gen takes up to 10-20 minutes, so i didnt know if there wasnt something i could change or add to improve things

[ornate elk]

What's your GPU and how much RAM do you have?

[mental vector]

@ornate elk, do you have a guide or general tips for setting up in WSL2? I have a Win11 with WSL2 running Ubuntu. My GPU is a Nvidia RTX 3060 12G, I specifically need the WSL to "see" the GPU device from the host.

After that, I'll install ComfyUI

[ornate elk]

Why do you need WSL for that?
Normaly a 3060 should show up without any issues

[mental vector]

The reason is that I don't want to fuss with Python in a Windows env.

However, if you think that's the best course of action, I'll give it a try.

[ornate elk]

There is nothing to worry about python on windows. If you dont have multiple projects requiring different versions of python your fine.

[mental vector]

Sounds good, Python 3.12 for ComfyUI is all I need. I wanted to try running it my gaming PC, I have another right with a less-than-stellar GPU that I'm using for image generation.

[ornate elk]

Ah okay, yea then just install the latest 3.12.x version and in the first screen check "add python to path"

[mental vector]

Thank you!

[fiery forge]

ah sorry i fell asleep not long after i sent that

ive got an amd radeon rx7800 and 37868 ram

[ornate elk]

Ah no problem.
For AMD you have to use my AMD setup Guide with Zluda. Either Auto1111 or Forge with Zluda).
With that you can generate images in seconds.
The Guide is on the first link of the pinned messages in this channel.

[fiery forge]

thanks

i followed the first guide i watched to a t but it was from last year, it would drastically slow down my pc and some images took 30 minutes

[tall pier]

hey so I have been using stable for a while now but how do I open this again once I close the terminal?

[ornate elk]

Hey, you have to run the webui-user.bat again to use it

[ornate elk]

Yea most AMD Guides are outdated and wrong. Thats why I made that guide list on github with updated and working setups

[peak kelp]

hello

im trying to use controlnet openpose

but its not working

idk whyt

[dire horizon]

hello what are some good recourses for me to learn fine tuning a diffusion model ?

[ornate elk]

Your image is already processed and so you have to set the preprocessor to None

If the input image would be a human then you have to set the preprocessor to Openpose then control will process it into a pose.

[tame cave]

Hey i have a weird question, is there a way to have info from an image (prompt, lora, models, cfg, etc...) i generated ? i generated a ton from last year and since i reset SDForge i don't remember which set i use...

what was that ?

holy fuck almost clicked on an invite... need an admin

[peak kelp]

i did some investigation then i realised i was using the 1.5 version in SDXL so i switched to SDXL controlnet model and everything started working

[ornate elk]

hey yes you can load them into the PNG-Info tab to get all the information back

[tame cave]

Alright was gonna be mad to lose some precious on some good generated images

Thanks a lot

[harsh token]

Does anyone here know how to keep Wan 2.2 from changing brightness and contrast? I'm trying to do a frame to frame to create a seamless loop, but the brightness especially just shoots up very intensely.

[delicate citrus]

hi guys, how can I use HF Downloader? I use Comfy on colab, so I neet to do the download directly

[delicate citrus]

why not? this processor image did you load or run it?

[delicate citrus]

you can use the exttension Civit AI Browser +, it download the models directly to you folders

and SD Hub to download extensions, controlnet, etc too

[tranquil siren]

Is this the are I could ask assistance in using ControlNet? Specifically I am trying to use it to modify poses of an image i generated so that i can maintain image continuity but in different poses

[cobalt agate]

@harsh token Did you change the default negative prompt? I think there are quite a few tokens in there to control brightness and color bursts.

[harsh token]

My negative prompt is empty. Is there something I can add there to fix the brightness/color changes? Cause I've tried so many things at this point. I hooked up a color match node and it didn't fix anything either.

[cobalt agate]

Check the link, #6.
https://www.instasd.com/post/wan2-2-whats-new-and-how-to-write-killer-prompts

[harsh token]

Thanks I'll give that a try and see if it helps

[lime pivot]

hello im new to comfyui and this scared the shit out of me, am i suppose to be worried lol

[harsh token]

This didn't help. The video still gets really bright

[solid dirge]

does anyone know how to fix this?

[ornate elk]

ultralytics got fixed, if you installed impact pack rn then your safe

[hot thorn]

anybody here ever trained LORAs on civitai? I made my first one and it doesnt look great, i think i messed up something in the settings

[lime pivot]

@ornate elk btw bro, do you know whats wrong with this? I’m copying a workflow (https://openart.ai/workflows/-/-/Lj54kkgJeEgWSvxadvnq) and it said I had missing node types, so I pressed "install all missing nodes." Then I restarted and refreshed, but I got the same error, it said I was still missing those same nodes. Then, when I ignored it and tried to run the workflow anyway, I got "prompt execution failed... node id 77"

[ornate elk]

cant help with that sry

[livid jetty]

how do i get comfy-ui to use my models on my usb drive

[terse siren]

Should I just give up all hope of running Flux Kontext on my PC? I have an NVIDIA GeForce RTX 2080 SUPER with 8gb of VRAM

It takes hours to generate a single image

wait....I got it

i think

I lowered the vram slider lol

[cobalt agate]

@livid jetty Check out the WIKI under Advanced (scroll down a lot). There is a yaml file you can edit to extend the path list where is looks for models. You may have to assign your USB drive a drive letter on windows, however.
https://comfyui-wiki.com/en/install/install-comfyui/comfy-cli

[fiery forge]

do you have a reccomendation for which one to use? i don't know if i did something wrong but when i try to open it after completing everything it just refuses to open, mentioning its primarily used by 3.10.6 python and something about torch not being able to use gpu

and when i tried the 3.10.6 one like it initially informed me it said something about pip not working

[ornate elk]

Would recommend trying Forge with Zluda

Its a bit more updated than Auto1111

Feel free to let me know if you have any issues and provide some full cmd screenshots

[fiery forge]

ight bet

[ornate elk]

it says rocm version NONE

have you installed hip sdk 6.2 ?

[fiery forge]

no, was i supposed to

oh okay i see what you mean

[fiery forge]

no this error is while trying to run step. 7

[minor totem]

Hi, I'm trying to set up stable diffusion but there are way too many guides and different kinds so I'm very confused.
I have an AMD 6900 XT running windows. What do you recommend to use for best performance? And is there a particular guide you'd recommend?

[ornate elk]

Ah okay then just proceed with the zluda setup

[fiery forge]

gotcha

[ornate elk]

Hey in the first link of the pinned messages you find my AMD setup guides.
Would recommend going with "Forge with Zluda"

[minor totem]

♥️

[modern hound]

We are building a custom front end for SD/1111 by using the API. We are testing models and prompts using the Webui and passing these settings into our code. The thing is that the exact same seed and settings (we checked many times for possible syntax mistakes), are giving slightly different results. We are working on this issue for many weeks now... Any tips?

[ornate elk]

Did everything worked?

[fiery forge]

im a litle stumped on step 6 for adding zluda but i just started on it

[fiery forge]

trying to figure out where this is at

%HIP_PATH%bin

or am i just supposed to manually type both in or something

[upbeat atlas]

Hi, I recently re installed windows and am doing the same on the same GPU. I'm following the guide, but after running the webui for the first time I get this error:

[ornate elk]

You have to add them manually. Check the screenshots link below that step which shows how its done

[ornate elk]

Hey, did you installed windows no back on your smaller drive?

[upbeat atlas]

There are 2 drives, one smaller for windows and a bigger one for my stuff. I'm installing the SD on my bigger drive.

[ornate elk]

And what's your GPU again?

[fiery forge]

do i follow every step in the link or just the ones involving adding zluda and hip path

[upbeat atlas]

Amd 6900 xt

[ornate elk]

Only watch the two screenshots with the red numbers

[ornate elk]

Okay, if your at Step 7, just follow the guide to setup zluda

[upbeat atlas]

Is it fine that the webui didn't launch?

[ornate elk]

Yep, I have to edit the guide to make that clearer, that it won't launch if zluda is not setup

[upbeat atlas]

Good to know, thank you 🫡

[fiery forge]

@ornate elk any idea what i did wrong?

[ornate elk]

Delete the venv folder and relaunch

The rest looks okay

@fiery forge does it launches now?

[upbeat atlas]

Hi, so I did the rest of the steps and I still keep getting that error saying torch not compiled with cusa enabled

[ornate elk]

Delete the venv folder and the .zluda folder inside forge and relaunch

That should fix it

[upbeat atlas]

🫡

[fiery forge]

yea

thanks to you

[ornate elk]

Perfect.
If its compiling, just wait and then try to generate an image.
Compilation will take a while

After thats its fast

[fiery forge]

no issues as of rn its just making me download a model

so i just have to deal with a slow first gen and then its good after the fact?

[ornate elk]

Exactly, one time only. Not at every start.

And that compilation only happens again when torch, zluda or the amd driver updates

[fiery forge]

got it

thank you

[ornate elk]

No problem, let me know if you have any other questions!
But I'm going to sleep now

[ruby yoke]

you should probably ask in the #🐝｜swarm-ui channel on or @jagged plover 's discord

[livid jetty]

Also I swear I've gotten better image generation on Normal Stable Diffusion than comfy-yo. Is this normal?

[steady flint]

Guys, can someone please help me change this?

The new ComfyUI favicon is so fucking ugly

It's ignoring the custom one I made to use in place of the the pyssss catgirl

[steady flint]

I replaced the icon files at assets/images/favicon_progress_16x16/frame_9.png but ComfyUI is still using the ugly yellow blue one

[lament mulch]

stability matrix

[wispy knot]

Hello I need some help 🙂

Am using Forge UI and it does not want to do png info to txt2img WITH the picture. The txt2img TEXT works.

[bitter pier]

Could you post the error message here?

[ruby tide]

hey! could anyone help me with setting up sdxl, kohya and lora. itd mean alot

[haughty pilot]

is this important or should I just ignore it

[lost ice]

is there no good way to do video with forge? can't find anything that isn't pushing comfyui

[lost ice]

i think i figured it out, pancho's reforge has it

[fiery forge]

im getting a warning from the cmd about low vram but unless im reading it wrong i should have plenty, any ideas?

[ornate elk]

Which settings (resolution, steps, checkpoint) did you used when you got the warning?

[fresh crystal]

I can't get Kohya to do anything. It opens but trying to get it to actually serve a function shoots weird impossible to google errors. I'm pretty sure it's a fundamental problem with the installation but I can't find any guide that works like it's supposed to. The ones that aren't 2 years old are just summaries of the ones that are 2 years old. I'm on AMD and Linux.

[mild blade]

does anyone have familiarity with the kling img2video api?

how come the results are so much worse through the api than through their offical website?

[shadow sleet]

hello, i am currently having a problem with torch on my amd gpu, as it keeps saying:
RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check
how do i fix this?

[ornate elk]

Hey, what's your GPU?

You need the specific AMD webui version.
Follow any of the Zluda setup guides from the first link of the pinned messages in this channel.

[lament mulch]

Can anyone help me, pls? I installed flux1.1 dev in comfy ui to run it locally. Which installed some 23GB. Got a working prompt to image model in it, but don't know how to run an image-to-image model in Flux1.1 dev.

[wispy knot]

yes

[wispy knot]

how do i do this

[wispy knot]

hello! You can open your webui-user.bat with a text editor, then you can look up the commandline_args and add it there 🙂

[ornate elk]

He shouldn't add it there because hes on AMD and then it will use slow CPU mode instead.

[ornate elk]

Can you explain that a bit more or show some screenshots

[wispy knot]

alr 🙂

[wispy knot]

yes

*** Error loading script: infotext.py
Traceback (most recent call last):
File "C:\Users\user\Documents\Ai\webui\modules\scripts.py", line 525, in load_scripts
script_module = script_loading.load_module(scriptfile.path)
File "C:\Users\user\Documents\Ai\webui\modules\script_loading.py", line 13, in load_module
module_spec.loader.exec_module(module)
File "<frozen importlib._bootstrap_external>", line 883, in exec_module
File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed
File "C:\Users\user\Documents\Ai\webui\extensions-builtin\sd_forge_controlnet\scripts\infotext.py", line 7, in <module>
from internal_controlnet.external_code import ControlNetUnit
File "C:\Users\user\Documents\Ai\webui\extensions-builtin\sd_forge_controlnet\internal_controlnet\external_code.py", line 7, in <module>
from .args import ControlNetUnit
File "C:\Users\user\Documents\Ai\webui\extensions-builtin\sd_forge_controlnet\internal_controlnet\args.py", line 53, in <module>
class ControlNetUnit(BaseModel):
File "C:\Users\user\Documents\Ai\webui\extensions-builtin\sd_forge_controlnet\internal_controlnet\args.py", line 120, in ControlNetUnit
def bound_check_params(cls, values: dict) -> dict:
File "C:\Users\user\Documents\Ai\system\python\lib\site-packages\pydantic\deprecated\class_validators.py", line 234, in root_validator
return root_validator()(*__args) # type: ignore
File "C:\Users\user\Documents\Ai\system\python\lib\site-packages\pydantic\deprecated\class_validators.py", line 240, in root_validator
raise PydanticUserError(
pydantic.errors.PydanticUserError: If you use @root_validator with pre=False (the default) you MUST specify skip_on_failure=True. Note that @root_validator is deprecated and should be replaced with @model_validator.

For further information visit https://errors.pydantic.dev/2.8/u/root-validator-pre-skip

---

C:\Users\user\Documents\Ai\webui\extensions\stable-diffusion-webui-two-shot\scripts\two_shot.py:397: GradioDeprecationWarning: unexpected argument for Image: tool
canvas_image = gr.Image(source='upload', mirror_webcam=False, type='numpy', tool='color-sketch',
*** Error calling: C:\Users\user\Documents\Ai\webui\extensions\stable-diffusion-webui-two-shot\scripts\two_shot.py/ui
Traceback (most recent call last):
File "C:\Users\user\Documents\Ai\webui\modules\scripts.py", line 545, in wrap_call
return func(*args, **kwargs)
File "C:\Users\user\Documents\Ai\webui\extensions\stable-diffusion-webui-two-shot\scripts\two_shot.py", line 398, in ui
elem_id='twoshot_canvas_sketch', interactive=True).style(height=400)
AttributeError: 'Image' object has no attribute 'style'

Also i cannot seem to get a brush in my latent couple

[ornate elk]

Can you send the complete cmd log?

[wispy knot]

yes

[ornate elk]

That helps to find the error but it seems two show extension has issues

[wispy knot]

because I want to be able to make masks you know and to use ratio... THIS is really hard to do with regional prompter, or at least it seems so.

[ornate elk]

Have you tried Forge couple extension? Its easier to use than regional prompter and also supports masks

[wispy knot]

what is this? 🙂

and where can i find it?

[ornate elk]

Two shot extension is outdated since two years and doesn't work.
So better delete it from the extensions folder.
Then run the Update.bat in forge and then install the sd-forge-couple extension from here:
https://github.com/Haoming02/sd-forge-couple

[wispy knot]

okay thanks

[ornate elk]

You can install it by copying the URL and in the extensions tab click on install from URL.
Then paste it there and hit install

[wispy knot]

will try 😄

and the controlnet one is also not working.. when i am pasting to txt2img, it is NOT using that image as a style for some reason ;c

[wispy knot]

yeah thanks 😄

[ornate elk]

Maybe your missing IP-Adapter models?

[wispy knot]

not sure, i just want to use openpose, no tutorial stated that i had to download it i believe

[ornate elk]

Oh then thats the issue

Controlnet needs so called Controlnet models

[wispy knot]

oh i got them

for Illustrious

[ornate elk]

Ah that should work

[wispy knot]

yeah but it does not work ;c

and it gibs errors

[ornate elk]

Feel free to show me the errors and how your trying to use it

[wispy knot]

it is all in the file, but i know it is a long long read 😄 so i will show you yes

C:\Users\user\Documents\Ai\system\python\lib\site-packages\transformers\utils\hub.py:128: FutureWarning: Using TRANSFORMERS_CACHE is deprecated and will be removed in v5 of Transformers. Use HF_HOME instead.
warnings.warn(
Using pytorch cross attention
Using pytorch attention for VAE
ControlNet preprocessor location: C:\Users\user\Documents\Ai\webui\extensions-builtin\sd_forge_controlnet\annotator\downloads
*** Error loading script: api.py
Traceback (most recent call last):
File "C:\Users\user\Documents\Ai\webui\modules\scripts.py", line 525, in load_scripts
script_module = script_loading.load_module(scriptfile.path)
File "C:\Users\user\Documents\Ai\webui\modules\script_loading.py", line 13, in load_module
module_spec.loader.exec_module(module)
File "<frozen importlib._bootstrap_external>", line 883, in exec_module
File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed
File "C:\Users\user\Documents\Ai\webui\extensions-builtin\sd_forge_controlnet\scripts\api.py", line 17, in <module>
from scripts import external_code, global_state
File "C:\Users\user\Documents\Ai\webui\extensions-builtin\sd_forge_controlnet\scripts\external_code.py", line 1, in <module>
from internal_controlnet.external_code import * # noqa: F403
File "C:\Users\user\Documents\Ai\webui\extensions-builtin\sd_forge_controlnet\internal_controlnet\external_code.py", line 7, in <module>
from .args import ControlNetUnit
File "C:\Users\user\Documents\Ai\webui\extensions-builtin\sd_forge_controlnet\internal_controlnet\args.py", line 53, in <module>
class ControlNetUnit(BaseModel):
File "C:\Users\user\Documents\Ai\webui\extensions-builtin\sd_forge_controlnet\internal_controlnet\args.py", line 120, in ControlNetUnit
def bound_check_params(cls, values: dict) -> dict:
File "C:\Users\user\Documents\Ai\system\python\lib\site-packages\pydantic\deprecated\class_validators.py", line 234, in root_validator
return root_validator()(*__args) # type: ignore
File "C:\Users\user\Documents\Ai\system\python\lib\site-packages\pydantic\deprecated\class_validators.py", line 240, in root_validator
raise PydanticUserError(
pydantic.errors.PydanticUserError: If you use @root_validator with pre=False (the default) you MUST specify skip_on_failure=True. Note that @root_validator is deprecated and should be replaced with @model_validator.

For further information visit https://errors.pydantic.dev/2.8/u/root-validator-pre-skip

mmhhh

i am missing modules for controlnet it says as well

*** Error loading script: controlnet_model_guess.py
Traceback (most recent call last):
File "C:\Users\user\Documents\Ai\webui\extensions-builtin\sd_forge_controlnet\scripts\controlnet_core\controlnet_union.py", line 6, in <module>
from sgm.modules.diffusionmodules.openaimodel import (
ModuleNotFoundError: No module named 'sgm'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "C:\Users\user\Documents\Ai\webui\modules\scripts.py", line 525, in load_scripts
    script_module = script_loading.load_module(scriptfile.path)
  File "C:\Users\user\Documents\Ai\webui\modules\script_loading.py", line 13, in load_module
    module_spec.loader.exec_module(module)
  File "<frozen importlib._bootstrap_external>", line 883, in exec_module
  File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed
  File "C:\Users\user\Documents\Ai\webui\extensions-builtin\sd_forge_controlnet\scripts\controlnet_model_guess.py", line 10, in <module>
    from scripts.cldm import PlugableControlModel
  File "C:\Users\user\Documents\Ai\webui\extensions-builtin\sd_forge_controlnet\scripts\cldm.py", line 6, in <module>
    from scripts.controlnet_core.controlnet_union import ControlAddEmbedding, ResBlockUnionControlnet
  File "C:\Users\user\Documents\Ai\webui\extensions-builtin\sd_forge_controlnet\scripts\controlnet_core\controlnet_union.py", line 12, in <module>
    from ldm.modules.diffusionmodules.openaimodel import (
ModuleNotFoundError: No module named 'ldm'

---

*** Error loading script: controlnet_sparsectrl.py
Traceback (most recent call last):
File "C:\Users\user\Documents\Ai\webui\extensions-builtin\sd_forge_controlnet\scripts\controlnet_core\controlnet_union.py", line 6, in <module>
from sgm.modules.diffusionmodules.openaimodel import (
ModuleNotFoundError: No module named 'sgm'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "C:\Users\user\Documents\Ai\webui\modules\scripts.py", line 525, in load_scripts
    script_module = script_loading.load_module(scriptfile.path)
  File "C:\Users\user\Documents\Ai\webui\modules\script_loading.py", line 13, in load_module
    module_spec.loader.exec_module(module)
  File "<frozen importlib._bootstrap_external>", line 883, in exec_module
  File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed
  File "C:\Users\user\Documents\Ai\webui\extensions-builtin\sd_forge_controlnet\scripts\controlnet_sparsectrl.py", line 7, in <module>
    from scripts.cldm import PlugableControlModel, ControlNet, zero_module, conv_nd, TimestepEmbedSequential
  File "C:\Users\user\Documents\Ai\webui\extensions-builtin\sd_forge_controlnet\scripts\cldm.py", line 6, in <module>
    from scripts.controlnet_core.controlnet_union import ControlAddEmbedding, ResBlockUnionControlnet
  File "C:\Users\user\Documents\Ai\webui\extensions-builtin\sd_forge_controlnet\scripts\controlnet_core\controlnet_union.py", line 12, in <module>
    from ldm.modules.diffusionmodules.openaimodel import (
ModuleNotFoundError: No module named 'ldm'

so maybe you are right after all

[ornate elk]

It says controlnet_union but thats not for illustrious

If you have that

Can you screenshot the webui with the controlnet Settings?

[wispy knot]

how do i do this?

[ornate elk]

Win+Shift+S

[wispy knot]

oh you mean just the control net xd

[ornate elk]

What is the openpose s6000

[wispy knot]

oh it is a model

[ornate elk]

Where have you downloaded it from

[wispy knot]

i dont know anymore xD

[ornate elk]

Thats probably the issue

Here is the correct openpose model for illustrious:
https://civitai.com/models/1359846/illustrious-xl-controlnet-openpose

[wispy knot]

yeah that is what i use.. it switched for some reason to that one -.-

[ornate elk]

Can you try again?

[ornate elk]

With that model

[wispy knot]

hahah it is not about that, it starts, but.. it doesn't work you kow, for some reason

i will try it again!

[ornate elk]

You dont have an image loaded in

In the screenshot

[wispy knot]

yes, it gives the arror when I start

[ornate elk]

Can you show me the error in the webui

Not cmd

Maybe a full screenshot of the whole thing

[wispy knot]

it does not show in the webui 😄

at all

only on the cmd

it does not give an error message.

[ornate elk]

When does the error in the cmd appears?

[wispy knot]

start-up

[ornate elk]

When pressing Generate or when loading in the image?

[ornate elk]

By starting the webui fresh?

[wispy knot]

correct

[ornate elk]

Ohh

Have you run the update.bat ?

[wispy knot]

yes

i did that the first thing when installing it 🙂

[ornate elk]

And you installed it today?

[wispy knot]

3 days ago

i was not so proficient at this

at all

[ornate elk]

Can you restart and show the me full cmd log again?

[wispy knot]

okay

[ornate elk]

And also please screenshot your content of the webui-user.bat

[wispy knot]

ok 😄

Already up to date.
Press any key to continue . . .

update

NOTHING done yet, just startup

[ornate elk]

Okay and the webui-user.bat?

I think the issue lies in a weird merge of Auto1111 with Forge that is done there with the -ref-home arg

You didnt installed forge the correct way

[ornate elk]

Thats the same log again

[wispy knot]

ohh really? Because i first had aaaa or something, it just.. takes info from the AAA stuff.. I could merge the files it said with this

[ornate elk]

Yea thats the problem

[wispy knot]

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--forge-ref-a1111-home "F:\stable-diffusion-webui" --cuda-stream --cuda-malloc --disable-gpu-warning

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

call webui.bat

[ornate elk]

Thats not recommended

[wispy knot]

OH

really?

i did not know

[ornate elk]

So did you installed it as Auto1111 at first and then switched to forge or do you have automatic still installed?

[wispy knot]

i found forge was better or something, so switched to forge. I have never used a1111

[ornate elk]

Okay but it lists a1111 as installed on the F drive

And forge is on C documents

[wispy knot]

yes

it is merged

[ornate elk]

Yep that broke it

[wispy knot]

so i need to change that?

[ornate elk]

Yes

[wispy knot]

didn't know wow

[ornate elk]

You should safe your models and outputs of boath and then delete boath and install forge the correct way.
Guide for the setup is in the first link of the pinned messages

[wispy knot]

do i need to completely reinstall 😮

or can i just.. delete the merge?

[ornate elk]

You can try remove the --forge-ref-a1111-home "F... Arg from here and then maybe it works again

[wispy knot]

yea

[ornate elk]

But if not then you need a clean reinstall

Also normaly forges folder has a different name if you download forge as zip file which is recommended

[wispy knot]

ok

i donwloaded it as a zip 😄

the all in one thingy

webui_forge_cu121_torch231(1)

[ornate elk]

Yea that

[wispy knot]

it still has errors

is it not just due to my python version?

[ornate elk]

Forges zip comes with it's own python

Normaly that shouldnt be an issue

[wispy knot]

just to make sure

[ornate elk]

Which error do you get?

[wispy knot]

3.10.6

right?

[ornate elk]

Thats okay

[wispy knot]

same as before

it works you know, but..

some features do not work

[ornate elk]

Can you share the log again?

[wispy knot]

can you share your log?

[ornate elk]

Can try to move forge to C

Documents folder is not recommended as that can cause permission issues

[wispy knot]

OH

and C better?

or maybe better to put it on a diff place alltogether?

[ornate elk]

Its fine if its in a folder you created like C:\ai\ or F:\ai\

Just not in documents, downloads, desktop, program files, or external drives

[wispy knot]

can i put it on an hdd or is it then slower?

[ornate elk]

It will be slower

Best is to have it on an SSD and later you can place the models on an HDD and link them to the webui

If you have not much SSD space
Then only the model loading time is increased but python can be fast on the ssd

[wispy knot]

yea

makes sense 😄

still gives errors

where is the clean install tagged 🙂

think i fucked something up

as as i also installed some shit on spaces

idk what it is

[ornate elk]

What exactly?

Or where

xD

The setup guides can be found here:
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides

But forge is just a all in one zip. It should work normaly.
Im at work rn and tested it on my pc and no errors here

[wispy knot]

backing up all the other stuff now 🙂

thanks

[ornate elk]

NP, let me know if it works

[wispy knot]

am back, was eating some. Running update.bat now

Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Version: f2.0.1v1.10.1-previous-669-gdfdcbab6
Commit hash: dfdcbab685e57677014f05a3309b48cc87383167
Installing clip
Installing open_clip
Installing xformers
Fetching updates for huggingface_guess...
Checking out commit for huggingface_guess with hash: 84826248b49bb7ca754c73293299c4d4e23a548d...
Previous HEAD position was 3f96b28 Update init.py
HEAD is now at 8482624 Update model_list.py
Installing requirements
Installing forge_legacy_preprocessor requirement: fvcore
Installing forge_legacy_preprocessor requirement: mediapipe
Installing forge_legacy_preprocessor requirement: svglib
Installing forge_legacy_preprocessor requirement: depth_anything_v2
Launching Web UI with arguments: --cuda-malloc --cuda-stream --xformers
Using cudaMallocAsync backend.
Total VRAM 12282 MB, total RAM 32722 MB
pytorch version: 2.3.1+cu121
WARNING:xformers:A matching Triton is not available, some optimizations will not be enabled
Traceback (most recent call last):
File "C:\Ai\system\python\lib\site-packages\xformers_init_.py", line 57, in _is_triton_available
import triton # noqa
ModuleNotFoundError: No module named 'triton'
xformers version: 0.0.27
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA GeForce RTX 4070 SUPER : cudaMallocAsync
VAE dtype preferences: [torch.bfloat16, torch.float32] -> torch.bfloat16
Installing bitsandbytes==0.45.3
CUDA Using Stream: True
C:\Ai\system\python\lib\site-packages\transformers\utils\hub.py:128: FutureWarning: Using TRANSFORMERS_CACHE is deprecated and will be removed in v5 of Transformers. Use HF_HOME instead.
warnings.warn(
Using xformers cross attention
Using xformers attention for VAE
ControlNet preprocessor location: C:\Ai\webui\models\ControlNetPreprocessor
2025-08-12 12:55:47,468 - ControlNet - INFO - ControlNet UI callback registered.
You do not have any model!
Model selected: {'checkpoint_info': None, 'additional_modules': [], 'unet_storage_dtype': None}
Using online LoRAs in FP16: False
Running on local URL: http://127.0.0.1:7860

To create a public link, set share=True in launch().
Startup time: 156.5s (prepare environment: 129.8s, launcher: 0.7s, import torch: 18.0s, initialize shared: 0.3s, other imports: 0.8s, load scripts: 3.1s, create ui: 2.5s, gradio launch: 1.3s).
Environment vars changed: {'stream': False, 'inference_memory': 1024.0, 'pin_shared_memory': False}
[GPU Setting] You will use 91.66% GPU memory (11257.00 MB) to load weights, and use 8.34% GPU memory (1024.00 MB) to do matrix computation.

looks good so far!

[wispy knot]

ay you fixed my errors, thank you man ❤️

idk what xformers and triton is though

[wispy knot]

also it works way faster now

[ornate elk]

Perfect! No problem 🙂

[ornate elk]

Xformers is a performance boost

The triton warning can be ignored

[mossy pagoda]

Okay, so I don’t know how this discord feels about us downloading Stable Diffusion locally but when I did I got the common runtime error: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check. I have an intel GPU (since that matters) is anyone able to help me?

[bitter pier]

Intel GPU is not usable in torch.

Only GPGPUs are availble for torch.

[mossy pagoda]

Well I did try to download a NVidia GPU but that did not work

[bitter pier]

Download GPU? What do you mean by that?

GPU is kind of device not some software.

GOt it?

[mossy pagoda]

Yeah I meant a NVidia driver but that didn’t work so does this mean I’m SOL

[bitter pier]

You have the Nvidia gpu on your computer?

I mean devices.

[ornate elk]

What's your exact GPU ? Check the taskmanager performance tab and there screenshot the GPU part

[mossy pagoda]

This is my GPU

[ornate elk]

And how much vram does it have?

Are you on a laptop?

[ornate elk]

@mossy pagoda SDnext can work with Intel Iris XE GPUs:
https://github.com/vladmandic/sdnext/wiki/OpenVINO

[mossy pagoda]

So should I delete the file for Automatic1111 that's currently on my laptop?

[ornate elk]

Auto1111 can't work with the igpu, only with the very slow CPU mode

But I dont know how much faster the iris XE will be

[bitter pier]

Awesome!

I've never heard of it.

Thanks bro.

[hot thorn]

does anyone have experience using wan2gp?

[rose stirrup]

Has anyone run into a similar problem using a LoRa and a part of the body just persistently breaks every time? I work with my own prompts, but I’ll even copy-paste the prompts from a sample picture I’m trying to duplicate and it still persistently breaks the character’s torso every time. I don’t know what I’m doing wrong anymore and am feeling impatient and frustrated because the picture is generating perfectly fine, and then breaks right at the end and I have to wait for it to finish making an already broken picture.

[ornate elk]

Hey, can you show an example?

[mossy pagoda]

Okay thank you, hopefully I can find a version where it is friendly for intel GPU’s

[mossy pagoda]

Okay so update, I have been told not using 1111 is my best course of action, someone recommended comfy so I will give that a try tomorrow since it has an installation guide that includes Intel Arc but since I have an integrated intel graphics card, I'm basically SOL.

[ornate elk]

Have you tried SDnext with the Openvino guide?
Its the only option for iris XE

[haughty pilot]

a few days ago I generated an image, yesterday I generated the same image with same seed and options (identical) and today with literally the same options it's different, what happened? Every single option is the same and the image is almost identical but not quite (like in previous generations) when I look at the ones generated before and the one generated today there's a tiny bit of a difference, again, all settings are exactly the same

[mossy pagoda]

Not yet because I have a long day with my job

[ornate elk]

Can you show me the images? And which webui did you used?

[heady pine]

Could someone please tell me how I can get Automatic1111 working with a 5080 Master graphics card, I keep getting errors and messages when I open the command window, I have tried for hours trying to figure out a way to fix it, but nothing has worked.

[ornate elk]

hey, check the first link of the pinnes messages for the setup guides. they also include steps for the 50 series

[heady pine]

None are for the 5080, that I could see

[ornate elk]

from the guide

[heady pine]

I'm sorry for asking a dumb question with the picture you have shown, but do I not download a python version, will this do it for me?

[ornate elk]

you need python before yes

[heady pine]

3.10.6?

[ornate elk]

3.10.11 would be better

[heady pine]

ok

[ornate elk]

here are the setup guides:
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides

and if you want to use flux or chroma in addition then you would need to setup forge and not auto1111

[heady pine]

Sorry, do I keep Use admin privliges when installing py.exe ticked, or untick it but tick Add python.exe. to path?

[ornate elk]

yes keep admin privileges ticked

and also tick add python to path

[heady pine]

Sorry for not saying so sooner, but thank you very much for your help. Without you sending me that link to a workaround, I would have been completely lost.

[ornate elk]

No problem 🙂

[lapis raven]

Hello.
I'm struggling to install it on WSL.
"ROCk module is NOT loaded, possibly no GPU devices."
I keep getting this error. Could it be that the latest version of the device driver isn't compatible?
Sorry if I'm wrong.
My GPU is an RX5700XT.

[topaz hull]

@lapis raven When did you notice that mate 👋

[lapis raven]

It was just a few hours ago.

[topaz hull]

Have you contacted the Mod

[lapis raven]

No, I hadn't really thought about it that much.

[ornate elk]

Why do you want to use WSL?
I would recommend using zluda.
You find the setup guides on the first link of the pinned messages

[cold echo]

Hey guys can someone help me out with a comfy UI issue

I'm trying to use a workflow API, since the result gets saved to the disk and then I have to read them it's taking too long (working with fast api btw)

Do you have a node or any other idea that can directly stream it to me

Pls ping with response

[inland portal]

Is there a way to run local webui with colab gpu via ssh?

Amazon sagemaker is hard for me, too hard

[sweet silo]

if I install forge i can just reuse all my old lora's and models right?

without having to do weird manual linking in windows

like drag and dropping the files in a forge folder

[sweet silo]

the drag and drop just works fine, this is super similar actually lol. Super easy, except my cuda is messed up because of my new gpu still which sucks

[brittle lantern]

@ornate elk a little question if ur not busy what's the recommend "base" vram range for image gen in general? 16gb or 12gb?

[ornate elk]

Delete the venv folder and relaunch. What's your GPU?

You may need to do additional steps if its an rtx50

[sweet silo]

so i switched from a 3060 to a 5060ti. And that has broken things, because I seem to need a newer version of cuda or something

[sweet silo]

yeah

[ornate elk]

12 should be minimum but more is always better.

[sweet silo]

I don't even see a venv folder rn lol

[ornate elk]

Then follow the forge setup guide from here, it has the needed steps for rtx50:
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides

[sweet silo]

but i think it's just rtx50 stuff

[ornate elk]

Forge doesn't have one, but auto1111

[sweet silo]

ah this seems great thanks yeah

i was wondering where to paste the commands

now i can use this

[brittle lantern]

I see so sum like 5070 will be ok or u recommend another GPU for that range?

[ornate elk]

If you go purely for ai then a 5060ti with 16gb vram would be better

[sweet silo]

hmm is this always supposed to work

| 0/20 [00:00<?, ?it/s]CUDA error (C:/a/xformers/xformers/third_party/flash-attention/hopper\flash_fwd_launch_template.h:188): invalid argument

I don't know why it's not working really

ill try using the webui user bat

[inland portal]

Anyone?

I wanna run colab a1111 as remote gpu to local a1111. How can i do that?

@ornate elk is it possible for that?

[ornate elk]

Can you show the full cmd log,m

[ornate elk]

Maybe with an extension but I dont think there is one.

[sweet silo]

i was trying stuff from reddit

i will look for that full cmd log after I reinstall everything from scratch

[sweet silo]

First I am posting the cmd log from launching cmd in the system folder. Performing all 3 of the commands I was supposed to do

this seems like the issue

I now should run the run.bat

Or should I wait first 🤔

I don't know why i'm getting these errors when it just works for others doing all of this

I am now running the run.bat again as per the instructions

[ornate elk]

That can be ignored

[ornate elk]

cmd log looks okay here

[sweet silo]

oh good, I performed the run.bat. It opens the ui and i then test a random prompt with some old model i had lying around and get a connection timed out error

[ornate elk]

what does the cmd shows?

[sweet silo]

uhhh i tried to ctrl a and it dissapeared

let me open it again

okay i dont know why pressing ctrl a makes it dissapear

i screenshotted it this time beforehand just incase

seems to be the main issue

[sweet silo]

i'll try to get the text for you though

maybe if I just use my mouse

okay i have text but it's not letting me upload lol

[ornate elk]

can you show the full cmd log?

[sweet silo]

uhh I thought this was the full cmd log

[ornate elk]

nope you can scroll upwards on the cmd

[sweet silo]

ah dang yeah i cant use ctrl a so that's why i missed it

one sec

[sweet silo]

there we go 😅

[ornate elk]

try remove --cuda-malloc from the webui-user.bat

then relaunch the run.bat

[sweet silo]

i'm on it

[ornate elk]

if that doesnt work you probably need to remove --xformers too

[sweet silo]

it didn't, i'll try that too

🫡

[ornate elk]

or if you want to use xformers try to install triton with this command:
.\python\python.exe -m pip install triton

[sweet silo]

hmm I got another connection errored out thing

[ornate elk]

how much RAM do you have?

[sweet silo]

I can try this if you think it'd work, what folder would i start the cmd from?

uhhh

like 32gb I think

[ornate elk]

like you did with the torch and xformers commands before

in the system folder

[sweet silo]

ohh i see, so would I put that after the -- cuda stream?

I just removed those two commands from the webui.bat

[sweet silo]

I can't tell if you mean I should cmd there or go in webui.bat again, sorry I appreciate you trying to help me out

I feel like I should cmd

but i just want to be sure

[ornate elk]

no not there

and then .\python\python.exe -m pip install triton

[inland portal]

Are the alternative without extension?

[sweet silo]

this is what happens when I use cmd there

[ornate elk]

only complete cloud solutions. so you run the gpu ion the cloud + the webui too. Like with vast.ai or runpod.io

[ornate elk]

okay then nevermind

then just launch the run.bat with only --cuda-stream as launch arg

[sweet silo]

ah right yeah i did that and got a error too I think, i'll do that again so i can show you

[sweet silo]

this is what happens

[ornate elk]

go into the system/lib/site-packages folder and there delete the two or one xformers folders

then run the run.bat again

[sweet silo]

🫡

oh wow it may be doing things

@ornate elkIt seems to work now, thank you!

I doubt I would have figured this out without you haha

[ornate elk]

perfect no problem!

[sweet silo]

now I can do stuff for dnd again

[inland portal]

Ssh colab?

So I'll use colab gpu than local gpu

[ornate elk]

have you checked out this?:
https://github.com/cagliostrolab/forge-colab

[ember raven]

Hey guys, i need help. I have an AMD RX6900XT GPU and i'm getting this error when attempting to launch it. I'm running it on Windows 10, i'm using the A1111 Fork for the AMDGPU from "lshqqytiger" and in the Webui user i'm running it with "--use-directml --opt-sub-quad-attention --no-half --disable-nan-check --autolaunch" commandline arguments. Any advice on how to fix it? thanks so much in advance

[ornate elk]

hey you installed it the wrong way by using directml

but you should use zluda and follow this setup guide: (Automatic1111 with Zluda or Forge with Zluda):
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides

Let me know if any step isnt clear

[ember raven]

sure, so to clarify, on the link you provided, i follow the instrutions for the A1111 with ZLUDA, not the DirectML?

[ornate elk]

exactly

Zluda is faster and works much better than directml. directml is currently broken anyway in the webui

[ember raven]

thanks! i'll let you know if it works

[novel owl]

hello comfy ui with Zluda have AMD driver update issue for image generation ?

[inland portal]

Okay that's good one ngl

A question though can i use the public gradio link? so that when i use the local a1111 it uses the colab as backend to generate stuff. Or it's purely run in colab?

[inland portal]

@ornate elk the colab doesn't work with me

[lyric vector]

for some reason when I got to male an image it dosnt generate it just pops up a grey box

idk what to do I tried everything in the note pad stuff to fix it but it isnt working

[ornate elk]

Hey, nope 25.8.1 seems fine so far

[ornate elk]

Oh okay :/

[ornate elk]

What's your GPU? And which webui do you use?

[lyric vector]

My is nividia geforce rtx

Mine*

When I first got this setup it worked perfectly but today its like it dosnt wanna work at all

So idk why I keep getting the grey boxes when I never got em before

[ornate elk]

Which one exactly and which webui?

[lyric vector]

A1111

[ornate elk]

Grey images mostly appear if someone tries to use a lora as model

[lyric vector]

Nope Im using models it worked befofe when I first made this deffusion thing but this time it didnt work

[ornate elk]

Can you show your settings used? Or the cmd log?

[ornate elk]

Can't help if you dont provide more information.
Gray images can be fixed easily normaly

[inland portal]

Wait i forgot to say it work after i mess with dependencies.

[ornate elk]

Oh nice

[inland portal]

And yep now i wanna know if it's possible to hook it to local. Like the api setting.

[ornate elk]

Idk but I guess not

[lime pivot]

Hey guys, sorry im really new and bad at comfyui, does anyone know what the problem with this missing nodes is? I was able to solve some other stuff by following Civitai comments + ChatGPT, but have not been able to solve this no matter what. GPT told me to rename the class in the py file in Note Editor to match the missing node, so it went from 'CombineTexts' -> 'Bjornulf_CombineTexts', but it didn’t work. It told me to do a bunch of other stuff that didn’t work too. #https://civitai.com/models/1631379/wan-21-vace-controlnet-workflow-from-youtube (the workflow)

[novel owl]

After an driver updtade i need to wait for re compile the API ?

[jagged grove]

Hello I am building an AI art focused computer with top tier parts. I have everything ironed out except my ram. Choosing between G.Skill Trident Z5 Neo RGB 96GB (2×48GB) DDR5-8000 CL38
G.Skill Trident Z5 Neo RGB 128GB (2×64GB) DDR5-6000 CL30

Any advice would be appreciated. My cost on parts besides this is around $7500 and ( wanted to get DDR5 8000 2x64gb sticks but they haven't released yet so I need to make a last minute call

[ornate elk]

Yep when torch, zluda or the and driver updates zluda needs to compile again

[novel owl]

i have a problem it wont work i have re install zluda and HIP sdk and put ZLUDA file on venv after delete it and now my comfy ui crash TuT

XD

again ?

you remember CS1o ?

[novel owl]

Checkpoint files will always be loaded safely.
Traceback (most recent call last):
File "D:\IA\SD-Zluda\ComfyUI\main.py", line 136, in <module>
import execution
File "D:\IA\SD-Zluda\ComfyUI\execution.py", line 13, in <module>
import nodes
File "D:\IA\SD-Zluda\ComfyUI\nodes.py", line 22, in <module>
import comfy.diffusers_load
File "D:\IA\SD-Zluda\ComfyUI\comfy\diffusers_load.py", line 3, in <module>
import comfy.sd
File "D:\IA\SD-Zluda\ComfyUI\comfy\sd.py", line 7, in <module>
from comfy import model_management
File "D:\IA\SD-Zluda\ComfyUI\comfy\model_management.py", line 195, in <module>
total_vram = get_total_memory(get_torch_device()) / (1024 * 1024)
File "D:\IA\SD-Zluda\ComfyUI\comfy\model_management.py", line 146, in get_torch_device
return torch.device(torch.cuda.current_device())
File "D:\IA\SD-Zluda\ComfyUI\venv\lib\site-packages\torch\cuda_init_.py", line 1071, in current_device
lazy_init()
File "D:\IA\SD-Zluda\ComfyUI\venv\lib\site-packages\torch\cuda_init.py", line 403, in _lazy_init
raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enabled

[ornate elk]

it seems like you need to install the updated ComfyUI-Zluda Version by PatientX

[novel owl]

i need to re install completly ?

[ornate elk]

yep because one fork of comfyui was outdated

[novel owl]

hawn

okay

[ornate elk]

Follow the "New Install Method" from here:
https://github.com/patientx/ComfyUI-Zluda/issues/188

[novel owl]

that wont work

same error

[ornate elk]

can you show the full cmd log?

[novel owl]

i have HIP sdk 6.2.4
i have install zluda on C:ZLUDA

okay

do you know what is the problem ?

[ornate elk]

Nope, whats your gpu again and have you done every step from the patienx new method correctly?

[novel owl]

RX 7900 XT

and the new method are marked problematic no ?

[ornate elk]

shouldnt be problematic for the 7000 series

[novel owl]

ah okay

[ornate elk]

have you restarted the PC after installing HIP SDK ?

[novel owl]

yes

[ornate elk]

do you have the zluda 3.9.5 files in C:\Zluda ?

[novel owl]

dont work

[ornate elk]

can you open up a cmd and run
zluda --version

[ornate elk]

hmm strange

tested my comfyui setup (not patientx ones) and it still works.
Im trying to setup patient x ones now

[novel owl]

do you want try to take controle of my computer for check if i forgot some thing ?

[ornate elk]

do you have the other comfyui version still installed too?

[novel owl]

yes

[ornate elk]

okay then i can check

[novel owl]

have you any desk ?

[ornate elk]

you should send ids or codes only in dm ^^

[novel owl]

i have a password

dont wory

[ornate elk]

ahh okay

[bitter pier]

This is the dependency issue and you can fix it by reinstalling torch.

[violet inlet]

what is the password? 🤣

[bitter pier]

The password of Anydesk.

[bitter pier]

Have you fixed?

[lyric vector]

Idk what else to say I tried to edit the stuff in the note pad to try and fix that and it still didnt work

[bitter pier]

Hi Can you show me how it is working now?

With some screenshots.

[ornate elk]

you dont need to say anything if you show more information, like screenshots of the webui-user.bat content or the full cmd log or your whole settings used in the webui like model and resolution etc xD

[lyric vector]

Ill do that later when I get some time

[bitter pier]

Sure. And I'm not so active here so never forget to ping me. Or DM?

[lyric vector]

Alright

[ornate elk]

Please send the screenshots here

[lyric vector]

Ok

[celest grotto]

Hello, I'm new here and I am currently trying to download and set up my own Stable Diffusion Automatic 1111 but I encounter this error. Can anyone please help me?

[ornate elk]

hey please follow the updated install guide from the first link of the pinnes messages

[prisma moss]

is it possible to downscale the image by x2 after x2 hrfixing ?

(forgeui)

[lyric vector]

heres a screen shot of this um what all did you all need

but if you all need any other screen shots let me know

@ornate elk

[onyx solar]

That hash is an Illustrious LoRA, not a checkpoint. You can't gen simply off of a LoRA.

You get the blue static like above.

This is what he's got selected:
https://civitai.com/models/440167/ragatha-the-amazing-digital-circus-ponydiffusionxlillustriousxl

[lyric vector]

so what else do I need cause I forgot

[onyx solar]

Choose an actual checkpoint file. Put that LoRA into the folder for LoRA files, then prompt the LoRA.

But also, become at least basically familiar with what you're trying to do so that you don't run into this kind of thing. Watch some videos about the basics. Learn what some of the terminology is. Read up on the best ways to gen images today. Pay attention to terms you don't recognize and look them up when you're downloading the files. All the basic stuff you should do when learning new things. I know it seems like I might be coming across a bit harsh, but like anything else, there's at least some learning to do when you're wielding tools like this.

[lyric vector]

checkpoint file ok

so are the checkpoints like the syles like dream shaper and pony xl and stuff like that?

well I tried using pony deffusion on the vae settings but I guess that didnt work

[onyx solar]

Pony is a whole other thing with its own complications.

[lyric vector]

oh ok

so what checkpoints work the best then?

[onyx solar]

For starters, I'd recommend getting checkpoints that are SD 1.5 or SDXL. If you do SDXL, change the resolution to 1024x1024, instead of the 512x512 you currently have set.

Once you learn some stuff, then go look at other types of models.

[lyric vector]

so how would I be able to use my loras on those?

[onyx solar]

I'd recommend not worrying about that until you know how to generate reliably without them.

That's like...step 2 in the learning curve.

[lyric vector]

ik how to generate stuff I just wanted use lora models now

[onyx solar]

Step 1 is generate reliably.

[lyric vector]

ok

[onyx solar]

Then just google how to use LoRAs with A1111

[lyric vector]

alright

[mossy pagoda]

Sorry for both the late night response but also the late response, currently in the process of using this and it is working for me so right now it is downloading and I am just waiting for it. Now I will have to figure out/look for a way to import custom Lora's from CivitAI

And I just got this wall of text

[onyx solar]

well, first, you are repeating the same parameter twice. Second, it's misspelled.

[mossy pagoda]

I did not put those in parameters in, it just happened and these were my packages

Just ran it again and now I got the error code once

[celest grotto]

I did it. Thanks a lot guys.

[celest grotto]

So, I did it and tried to generate the image. However, it is extremely slow and the picture didn’t turn out very great. Can anyone give me an idea on how to improve?

[ornate elk]

What's your GPU and which webui do you use?

[ornate elk]

The warnings can be ignored. If the ui launches normaly

[mossy pagoda]

The UI did not launch at all

[ornate elk]

Oh okay xD

Have you removed the two arguments?

[hot thorn]

Hi, i finally upgraded to 32gb of ram, and I know i set some things in my webui-user to allow me to utilize 16gb without hiccups. but now that I got 32gb, with browser on and AI running it's using 17/32 gbs, what should I get rid off in my webui-user to make AI make full use of more ram i got?

[ornate elk]

Ram is only needed for storing models and moving them to the GPU when generating

[hot thorn]

ye, but I had an issue where running SD and opera (where its open), uses 95-98% of my ram

[ornate elk]

You can increase the model count that you can store in RAM if you want to quickly switch between them

[ornate elk]

Yea 16gb is to less for that

[hot thorn]

To less for that? I dont understand

[ornate elk]

With the 16gb RAM and sdxl based models the ram gets full quickly

But with 32gb your fine

You dont need to change anything

[hot thorn]

Ok

By any chance, are u knowledgeable about hardware?

should i ask here? or should I dm u since its not SD related

[ornate elk]

You can ask here too

[hot thorn]

I have a weird ram situation, till today i had 2x 8gb ram sticks by hyperx fury, model: HX432C16FB3AK2/16, and today I installed 2 more 8gb sticks model HX432C16FB/8. my old ram could run XMP and it showed up in bios for me to choose the profile. However with the new ram sticks added XMP option is gone. Both ram stick sets support 3200mhz, so there should be a way for me to increase them, but there's no xmp options. using CPU-z I can see that my old ram looks like the first screenshot, while the new one looks like second screenshot. I'm using Aorus B450 Elite motherboard and ryzen 7 2700x cpu. What can I do? I already tried resetting CMOS, to no avail, and I swapped my original ram sticks from slot 2/4 (new in 1/3), to 1/3 and new in 2/4, but it also didnt help, and in my system I see they're stuck at 2400mhz

[ornate elk]

You probably need to do a BIOS update then. They often add better support for 4x RAM slots used.

[hot thorn]

I had a buos update some half a year ago. Where can i check bios version?

[ornate elk]

In cpu-z under mainboard

[hot thorn]

says my bios is F67d date 09/02/2024

[ornate elk]

Hm okay there is only one newer version for it which only adds a security fix

Nothing ram related

Also have you checked that the RAM is in the compatibility list of the Mainboard?

[hot thorn]

ye, it is

i manually changed to 3200 and it locked itself. changed to 3000 and it works

by locked itself i mean it just kept on resetting itself

[ornate elk]

ah okay so 4x 3000Mhz works now?

[hot thorn]

ye

i dont mind going from 15gb 3200 to 32gb 3000

[ornate elk]

yea thats not an issue

i had the same problem on DDR5 a few months ago but there 4x 6000mhz doesnt work. only 4x4000 and thats bad

[hot thorn]

if during generation ram was going all the way to 95% usage, now that i have more ram it should work slightly better?

ye 4k and 6k is a big drop

[ornate elk]

yea you can also check your pagefile settings and set them back to system managed for C.

because we mostly increased it to run better with only 16gb of ram

[hot thorn]

will that help a lot? or not much

[ornate elk]

it will help more probably

not much but the ram usage should be managed by ram and not ssd cache mostly

[hot thorn]

Can u remind me how to change it?

[ornate elk]

sure scroll down here for the steps:
https://www.tomshardware.com/news/how-to-manage-virtual-memory-pagefile-windows-10,36929.html

just make sure its on system managed for C and not custom. and that its only enabled for C drive at all.
Then reboot if you changed something

[celest grotto]

Nvidia with 6GB dedicated memory. I use Stable Diffusion Automatic 1111 webui

I did use ComfyUI and it worked just fine and very well

But as a newbie, ComfyUI is hard to use

[ornate elk]

ah okay, then edit the wbeui-user.bat
at the line set COMMANDLINE_ARGS= you need to add --medvram --xformers
then save and relaunch and it should be much faster

[celest grotto]

Thank you. I try but it is still taking 43 mins for one image. Is that normal?

[ornate elk]

absolutely not, on your gpu it should take not longer than 1-2mins

can you share the cmd log ?

and maybe the resoltuion and checkpoint you used

or screenshot the whole txt2img site

[celest grotto]

It takes a bit faster now with 8 mins generation times.

But 1 - 2 mins?

That's surprising

[prisma moss]

you shouldve gone 2x16 instead

adding 4 sticks on 80% of motherboards stresses the memory controller on it, hence it may cause instabilities, or a not working xmp profile (or lower speeds)

[hot thorn]

If i had more money to spare i would, 2x8 gb was £40, 2x 16 is 120-140

and im buying a gpu on top of that

[prisma moss]

did you consider selling old kit

idk where u r from but 2x16 3600mhz of kingston memory costs 80eur here

[hot thorn]

idea was to get 32gb ram, get gpu and in about a year just sell the whole setup, then build something actually around 2025 standards

Kingston Fury beast, new is £125 in uk

[hot thorn]

sent u a friend inv, i was in your situation sometime ago, and used the same model, can help u out in dms

[prisma moss]

r u on a1111 or forge

[prisma moss]

im not even from the uk and it took me 5 seconds to find it cheaper mate

[hot thorn]

where do you check those? i just go to ebay and look for cheapest not broken ones that dont look like chinese knock offs

[prisma moss]

bruh, its new

[hot thorn]

ye, i read that, what did you use to find it

[prisma moss]

google

https://www.idealo.co.uk/compare/201483813/kingston-fury-beast-32gb-dual-kit-ddr4-3600-cl18-kf436c18bbk2-32.html

[hot thorn]

for some reason this listing didnt show up on my search wtf

[prisma moss]

idk

https://www.ebay.co.uk/itm/127304252535?var=0&mkevt=1&mkcid=1&mkrid=710-53481-19255-0&toolid=20006&campid=5337770566&customid=YnUbqLzLyl30PUeZQjH0Pw

heres the direct link

[lapis raven]

Sorry I'm late
I saw that wsl is faster
ROCm version

[ornate elk]

not by much currently

zluda is easier to work with on windows

[ornate elk]

its still not using xformers

you need to dd the launch args --medvram --xformers to the webui-user.bat to make it faster

[strong path]

does anyone know where the setting for increase the meter fron 2 to 6 or higher?
i did some find where then change but, they did not accept the change.
Is a Extension of Addiontal Network

[ornate elk]

Lora Weights mostly go from -2 up to 2
Anything above can deep fry the image output

[strong path]

Yes, but the some of lora that can above 2 using or below -2

[ornate elk]

Also the additional network extension never got an update since 2 years.
I'm surprised it even works xD

[ornate elk]

Can you show me one? Would like to test

[strong path]

one of lora i something use would be skirtlengt or lgirl that make a child to adult i will send link on site with i mean

[haughty pilot]

sorry, was busy, can I send it here or dm?

[ornate elk]

if its nsfw then dm

[haughty pilot]

it's not, I will send it here

it's all generated one after another, same seed, prompt and every single setting

all have slight differences (1 actually got a different hair color)

[ornate elk]

the image doesnt contain the metadata so i cant check/compare them

[wary vale]

Am I able to use Stable Diffusion to make image to gif (while keeping the transparent background)?

[lapis raven]

I'm an AMD, is that right?

[vocal burrow]

debian t shirt ?

Did you update anything in between images ?

[lapis raven]

Thank you.
I'll install it on my end.

[haughty pilot]

the shirt was supposed to be black but I'm assuming it got influenced by "spiral eyes"
those generations were back to back on the same seed without changing anything

[prisma moss]

that because every other image gets a +1 to seed technically, when done in a batch

so really you should be doing 1 by 1 if u are doing the same seed

[haughty pilot]

it wasn't a batch

separate generations

all had the same seed

[prisma moss]

Then some variance is expected

Depends also how good you prompted it, or how loose the tags are

[ornate elk]

Let me know if any step isnt clear

[lapis raven]

It was giving me an error until just now, but after some trial and error, it started up.
Looks like I messed up the startup options lol

I was running it on a different SD ...

[lapis raven]

There is an error related to dlm...

The message is this...

[ornate elk]

Hey don't follow the directml guide. You need to follow the Automatic1111 with Zluda guide from here:
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides

[lapis raven]

I did it according to the pin above, but maybe it's been updated...?

[ornate elk]

But in your cmd log it shows that you installed the directml version and even used the launch args for directml.

[lapis raven]

Could it be that something I installed in the past is causing problems?

[ornate elk]

So you can delete that webui version and restart by following the guide with Zluda

[ornate elk]

Have you installed the webui before?

Maybe you launched the wrong version then

[lapis raven]

I see.

By the way, what is the UI mentioned in the installation guide above?

[narrow bolt]

Hi everyone! I hope someone can help me with these two questions about native VACE nodes:

1. Can we replicate END and START % effect with native VACE?
2. Can we chain multiple native VACE nodes?

[ornate elk]

Automatic1111, but you could also install Forge with Zluda.
Boath have the same UI

[lapis raven]

If you're going to put it in, is it best to put it in the top 1111?

[ornate elk]

I dont understand the question sry

[lapis raven]

Is there a difference?

[ornate elk]

they are diffierent webuis. they offer different features and have a diffeerent usability.
With a 5700xt i would advice to either use Auto1111 with Zluda or Forge with Zluda.

They should give you the best performance/usabilty

[lapis raven]

Thank you. I see.

[lapis raven]

The same error occurred...

[lime pivot]

@ornate elk hey brooo im planning to switch from python 3.12 to 3.11.13 i was wondering if there will be problems? and if i have to redownload stuff? like pytorch or other softwares/files or anything really

for comfyui

[ornate elk]

can you show the cmd log?

[ornate elk]

hey, on windows you need python 3.11.9 64bit as that version is the last one with an install package.
if you downgrade you only need to delete the venv folders of all ai tools that used the 3.12 python.
Then relaunching them rebuils the venv with the python stuff for 3.11

[lime pivot]

@ornate elk alright thanks bro!

[lapis raven]

I tried again and it started up without any errors... Thank you.

[ornate elk]

alright, np 🙂

[lapis raven]

It ended up recurring...😭

[carmine ibex]

Couldn't get Pilid nodes working so I auto updated comfyui and python and comfy no longer opened. Got this message: raise AssertionError("Torch not compiled with CUDA enabled") AssertionError: Torch not compiled with CUDA enabled

[ornate elk]

wrong launch args lol its not using zluda at all

it seems like you followed again the directml guide and not the one with zluda

is your webui installed in C:\AI\ or on D:\SD\ ? or do you have two installations maybe?

[lapis raven]

D:\SD\ only

It started up after I corrected the arguments.It started up after I corrected the arguments.

[ornate elk]

you need these: --use-zluda --update-check --skip-ort --skip-python-version-check

and you need the custom gfx files (gfx1010 xnack) for your gpu and HIP SDK 6.2 installed

everything explained here:
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides#amd-automatic1111-with-zluda

[civic fractal]

Using Wan2.2 I get this 9gb allocation runtime error when upscaling/interp 4x (doesn't happen on 2x) should I try winging it by adding this spare stick of mismatched 16gb RAM I have for a total of 48GB RAM?

[lime pivot]

just wanted to share but if you cant install wget for certain comfyui nodes you should upgrade to python 3.13. was trying to downgrade to 3.11 to download wget but "python_embeded\python.exe -m pip install wget" works for 3.13

[lyric vector]

Dose anyone know if I can make videos on this and if so how do I go about i

It*

[lime pivot]

@lyric vector comfyui

wan 2.2 or wan 2.1 if you wanna control the exact movement

[bitter pier]

Hey

Why you blocked me?

[carmine ibex]

Did a fresh install of comfyui with the installer. But I dont know where to view the terminal to see ongoing process and logs

[main rune]

if it's in the ui then this I guess? or are you talking about logs when it fails?

[carmine ibex]

I mean the black windows that comes up and shows everything. Like when you type in cmd

[main rune]

oh that

installer shenanigans I suppose, what did you use

[carmine ibex]

Googled the first thing that came up. I think ill delete it and do it manually from github

[lyric vector]

How do I set this up to do wan video cause it just automatically put me on images

[onyx solar]

Open up the workflow templates, click on video, find the wan one, download the models, mess with it, enjoy.

[onyx solar]

Click on the Workflow menu.

Not the sidebar.

[lyric vector]

Is this the menu or no

[onyx solar]

That is the menu, yes.

[lyric vector]

Oh ok I found it

Thank you

So I need all of these?

[onyx solar]

yup

[lyric vector]

I put the files into there correct folders and now I keep getting errors

Oh nvm guess I didnt

[main rune]

just been reading through the past chats so I assume that the latest radeon drivers can now work with zluda @ornate elk ? I'll try updating and troubleshooting regardless but this sounds like good news for me

[lapis raven]

Next, even though it can be started, cmd whispers, “Press the button...”

[lyric vector]

I tried tiwce to gen a video but it gets to this one section on the list of stuff it gose thorugh then its like it gets stuck there then this happens

[main rune]

...prging cache and rebuilding venv right now...

ah shit, gonna have to look up the full reinstall guide, was hoping it would just work right off the bat

well, I was overdue on updating everything anyways

should I unistall my old HIP drivers first?

[heady pine]

The prompting-help channel seems inactive, so I thought I might ask here. Does anyone know what I can do to fix this blur error with the head/face, it mainly happens when the head/face is a big part of the image, it doesn't really happen in other images.

[main rune]

shit, there's no support for gfx1030 files yet

should work fine but man

[main rune]

shit man

ugh I really hope I didn't accidentally missed something super specific like 'this install method is only valid for linux'

figured out the nccl part, windows was blocking the file for some reason, but now python is crashing

ah man...

oh crap, I might have forgot to update python

[ornate elk]

Yes the latest drivers are supported

And python 3.11.9 64bit needed

[ornate elk]

Yes to not get any issues

[main rune]

ah, but I run rx 6600 xt

python is still crashing, not sure what else I need to do, not sure if it got installed at all

[ornate elk]

Yea thats gfx1032

[ornate elk]

Which webui do you try to use?

[main rune]

comfy

[ornate elk]

Ah okay

[lapis raven]

I finally got it to start up...

[wary vale]

I'm in the process of learning SD. I am about to fetch the SDXL, but it asks me for a liscence on Stability.ai...

Do I need a liscence? I don't understand where to find this for personal use.

[ornate elk]

Does it work now?

[ornate elk]

If you download the files from stability ai on huggingface then you need to be logged in.
Would recommend just downloading the models from Civitai.com

[wary vale]

I have RTC GeForce 5070 Ti, and I'm running into problems with Python

My GPU is not compatible with SDXL? I have high end hardware...

This is so confusing I'm losing my mind, someone please help me

[ornate elk]

Check the first link of the pinned messages.
There follow the Automatic1111 or Forge setup guide.
Then you can use sdxl and other models

[wary vale]

I've been at this for 4-5 hours now, my head is exploding. I've been following a guide on YouTube, and got the assistance of GPT to do something with torches to fit my GPU. I hope the pinned messages helps.

But in theory, my RTX 5070 Ti works well with SDXL, correct?

[ornate elk]

Dont ask chatgpt for that stuff.
The guides in the first link of the pinned messages are updated

[ornate elk]

Absolutely

[wary vale]

After taking advice from GPT and tinkering a lot in CMD, should I perform a clean install of python?

GPT is drunk. I did a clean install...

Thank you for providing me the guide. Appreciate it.

[novel pendant]

Hello, let's continue here

because there are some problems with my discord account and I can send limited amount of messages in DM

[wary vale]

Add me as a friend and it should be sorted.

[novel pendant]

Ok

[wary vale]

You've just saved me a ton of headache.

[novel pendant]

Hello, I'll tend to collaborate with you on your project.

I thought about your suggestion carefully and I can do it

[wary vale]

If you struggle to send DMs on Discord, add me on TG as stated in DMs. Let's continue there, I can't talk too much about projects publicly yet. Especially not in the tech-support channel, it would clog.

[novel pendant]

Ok

[wary vale]

Regarding this. It's enough to open the cmd via Win+R 'cmd' - correct?

[ornate elk]

no

[wary vale]

Can you please guide me?

[ornate elk]

first issue is you downloaded the webui as zip

thats not how it works

you need to git clone it

like stated in the guide

[wary vale]

Okay, will do it immediately. I don't have to re-do steps around Python correct?

[ornate elk]

you need python 3.10.11 64bit installed

[wary vale]

This is the one I need?

Just double checking

It's already installed

[ornate elk]

yea now do the step where you click into the file explorer bar and type cmd and hit enter

then use the git clone command to download the webui

[ornate elk]

dont use that xD

[wary vale]

I'm at a loss, excuse my stupidity

What do you mean by File Explorer bar?

[ornate elk]

before

after

success

[wary vale]

Okay now I launched the webui again.

[wary vale]

Is this normal first execution?

[ornate elk]

nope, can you show the full cmd log?

[ornate elk]

ah you have to do the addiontal steps for the rtx

[wary vale]

Do I execute first, then do the additional steps?

Is that okay?

[wary vale]

Would you recommend ComfyUI or SwarmUI?

[ornate elk]

comfyui probably, but just for images, forge

[wary vale]

I got this working now thanks to you (thank you very much...). The next step is to fix ComfyUI/Forge I guess?

[ornate elk]

do you need comfyui?

and no problem

[wary vale]

I want to be able to create the objectively best images

Even if advanced, I am going to start diving into it all

Right now I am not very impressed by the images it produces haha

[olive terrace]

I tried many times, but I didn't get any image results. Why is that? Are there too many requests? But my credit keeps getting used up.I used n8n

[ornate elk]

please contact stability by email, this is not an official support channel for their products

[ornate elk]

because you need to learn how it works xD that takes time

[wary vale]

I got time

[ornate elk]

comfyui offers the most settings/features but is the hardest to get into

[wary vale]

But I guess I'm not finished yet given I don't even have ComfyUI yet

[wary vale]

But in general, ComfyUI is the one you recommend?

[ornate elk]

depends on what you want to do

for starters i dont recommend comfyui

[wary vale]

Is it better to learn Forge first?

And then advance to ComfyUI when I'm more comfortable?

[ornate elk]

yes so you learn the basics and how everything works

[olive terrace]

what is the email

[ornate elk]

mb they only have a ticket support here:
https://kb.stability.ai/knowledge-base/kb-tickets/new

[limpid loom]

So, I managed to get SDNext to run using ZLUDA, but when I tried to generate an image, I got this:

new() received an invalid combination of arguments - got (Tensor, requires_grad=bool), but expected one of: * (*, torch.device device = None) didn't match because some of the keywords were incorrect: requires_grad * (torch.Storage storage) * (Tensor other) * (tuple of ints size, *, torch.device device = None) * (object data, *, torch.device device = None)

How do I fix this? More importantly, what file do I need to look for in order to fix this issue?

[ornate elk]

Can't help with sdnext but would recommend forge with Zluda instead

Much easier to setup and work with

[chrome tangle]

Sorry for bothering, And CS1, sorry i haven't been around. I have been having some... Pet issues. Anyway. i am back now and i hope you are okay. And as for a new questionb. My friend told me to look into SPDA for my 5080. But they went to bed withotu explainign what it is playfully shakes fist . Is there a quick guide of what it does, any risks and if i should install it.

[ornate elk]

hey, which webui do you use?

[chrome tangle]

Hey, i hope you are doing okay! And i use Forge Via Stability matrix. But i can use the base version of forge.

[ornate elk]

i guess he means that you should use sdp attention as performance boost. you can enable it in the settings of the webui at the performance settings

[chrome tangle]

where is that tap if its okay to ask. Also, is there any risk to using it?

[ornate elk]

its under settings -> Optimizations -> Cross attention optimization (select sdp attention)
and there is no risk

[chrome tangle]

Thank you so much!

[ornate elk]

no problem

[chrome tangle]

i am guessing i need to install it first

[ornate elk]

ah nope then its not available in forge

then just make sure you have--cuda-stream as launch arg for forge

[chrome tangle]

Okays, thank you ❤️

[main rune]

python just crashes when I try to start comfy, any ideas?

[ornate elk]

Which Python version?

[hot thorn]

are images generated in normal SD and forge differ? I tried recreating a picture a friend sent me, with the exact settings, but it appeared different. quality was a bit lower and composition was different. I use reForge, and I think he just uses automatic1111

image generated by me has this at the end

while his has this

[pale garden]

Im getting a pytorch compatibiltiy error with forge. My gpu is 5070 ti, can someoen tell me what to do to get the right version

[ornate elk]

Hey, check the first link of the pinned messages. There follow the Automatic1111 or Forge Guide. They include the steps needed for the 50 series

[young cedar]

hi

i need help

i downloaded flux dev krea f16
and i dont know how to run it
please help

[main rune]

I think the latest, the guide claims higher is fine

[tired python]

Am I doing this right? The lora says clip skip 2, but I am unsure if it is talking about the clip skip and/or strength clip. It says strength .8 but again, I am not sure if it is talking about Strength model or strength clip.

[ornate elk]

not my guide then xD

3.13 is not recommended

[main rune]

Sigh

[ornate elk]

hey whats your gpu?

[tired python]

Can I get some help? I thought I did everything right, but I am consistently getting bad faces

[wary vale]

Is it difficult to set up an upscaler and sharpener of images on SDXL? Any updated guides on this?

[main rune]

it shouldn't be, you might need to download some upscale models but they're pretty straightforward

[main rune]

not exactly the best channel for this, there's #📝｜prompting-help . You should look up prompting guides for the model you're using, but I see you're using pony which needs quite a lot of prompting direction. You might want to try the newer models instead like Illustrious or NAI

otherwise I can only assume you're on too high CFG

[tired python]

Which illustrious do you recommend?

[main rune]

I've had good results with hyphoria, an Illustrious/NAI merge, but YMMV

[tired python]

Okay, my mistake using pony. I am using it with AI roguelite which is 100% normal sentence prompts

[mossy pagoda]

OpenVino does not work with my Intel IGPU laptop, am I cooked?

[tight flower]

are u looking for dev?

[celest grotto]

Thank you replying but I managed to do it. A guy on here helped me go through it.

[nova locust]

Hello, sorry for the question, but I'm new to this and don't know where to start. I've watched so many videos and tutorials, but not all of them are up to date, so I'm a bit lost. I'd like to generate all kinds of images on my local PC. Any advice or guidance on where to start?

[vocal burrow]

check up the pinned guide in comment and follow the setup for forge, it s the best compromise of easy to setup and powerful at the moment IMO.

[nova locust]

Thank you

[tall pier]

why am I getting all grey gens when using flux

is there a way to set it up?

[main rune]

What UI are you using, I think that Automatic1111 does not support flux models, but I am not sure

[vocal burrow]

yes auto doesnt support flux

but we d need more infos to properly troubleshoot the problem, what s the hardware, settings, logs

[cloud abyss]

I am having a problem when creating an NSFW image.
The image generated is gray.
Model downloaded from civitai.com and use of keywords.
For your information, my stable diffusion is CPU only.

[tall pier]

yes I am using A1111

now what

like do I need a different software to run it?

also I am using M1 mac

[vocal burrow]

no filter in local mode unless you activated it, not gonna discuss NSFW furthere here.

[vocal burrow]

Then yes you need something else, forge, comfyui, swarm, etc.

SD3 Large is the limit for A1111 , it can t even do SD3 Medium out of the box.

[tall pier]

is there a guide for mac

[vocal burrow]

cf the second pinned message

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

[cloud abyss]

Okay, thank you for your reply.
I will try to find out more on my end.

[vocal burrow]

It s likely that there is useful informations in the log

Try out other models to see if the problem is model specific

i m okay discussing technical stuff, you do you on your pc I don t care much about that. On a discord level I just won t discuss nsfw stuff here and on a personal level I m not helping with face transfer technology of any kind.

[hot thorn]

Can anyone tell me how much faster my image generation going to be after swapping to rtx 3060 12gb from gtx 1070 if my generation time was 6 min?

[cloud abyss]

Your generation will go from about 6 minutes to ~1.5–2.5 minutes per image, depending on the model and settings.

[vocal burrow]

you can compare with numbers from the community benchmarks https://vladmandic.github.io/sd-extension-system-info/pages/benchmark.html

[main rune]

I finally got to the ui screen!

[lapis raven]

Are there any command arguments for Radeon 5700XT?「 --use-zluda --update-check --skip-ort --skip-python-version-check --medvram --no-half --listen --port 80」

[main rune]

@pure tusk got a bot

[vocal burrow]

all cleared

[signal mesa]

Could someone please help me? I installed Deforum, but I get a warning that ControlNet isn't working. I installed it on my computer, but I don't know if I did it correctly. I copied all the files from the ControlNet folder to the folder where I have all the Stable Deffusion files. Is what I did correct? Why doesn't it still work?

[supple cape]

Is anyone else having triton language missing for ksampler advanced, I have Triton-windows installed but it's still complaining

Trying to get wan 2.2 i2v workflow working

[ornate elk]

Looks good maybe --opt-sub-quad-attention can help too

[ornate elk]

The controlnet models go into models/controlnet

[main rune]

GODDAMNNIT IT WAS WORKING FINE YESTERDAY

[lapis raven]

I got an error...

[thin charm]

Hi

[lyric vector]

I need some help i go to generate an image on stable deffusion it generates but I go to do another one and the thing craps out and just pops up a white screen

No thx thats just another scam

[vocal burrow]

@lime pivot this is a scam, get out of their discord

[rose apex]

I am trying to install A1111 Zluda on Windows. I have a 9060 XT 16 and I keep facing this problem where it says Rocm agent not found or something and I get a bunch of errors after that.

Has anyone been able to run Zluda with 9060 XT yet?

I have tried everything and it feels literally IMPOSSIBLE to fix

[ornate elk]

Can you share the full cmd log?

[ornate elk]

Can you show the cmd log?

Also which hip SDK version have you installed?

[rose apex]

I deleted that and am trying comfyui currently. I will send it to you when I try that again.