🤝｜tech-support

[slender sable]

Hi all, just wondering if I can get some help with a fresh install that is only creating grey boxes, the only model installed is Stable Defusion 1-3 and 1-4 and nothing else. I am not sure how to troubleshoot this.

I have a 4080 GPU and I have tried some fixes I found on line such as adding --xformers --medvram-sdxl --no-half-vae to the command line arguments but it doesn't seem to change anything.

[ruby yoke]

are you running intel or amd?

[slender sable]

AMD

5900x

[ruby yoke]

comfirm for me that you've gone through the AMD guides pinned at the top of this channel, please

[slender sable]

Sorry, I am using NVidia 4080, I thought you mean't CPU. If i follow the guide and go to NVidia automatic1111 it gets me to the below page which just loops back on itself. https://github.com/CS1o/Stable-Diffusion-Info/wiki/#nvidia-install-guides

[ruby yoke]

i did mean the CPU - I think you might want to wait till @ornate elk logs back on. he shouldn't be offline for too many hours

[quasi ginkgo]

Sup yall
Just made the jump from A1111 to Forge. Haven't genned much with it, been messing around with the settings and installing extensions and what not. Theres a setting that doesnt seem to be working for me though. I want to hide all incompatible lora/text invs for the current model, but whenever I save and reload the UI, it doesn't do anything

I didn't want to copy my huge ass models folder tho so I made a symlink for Forge to use my A1111 models folder

Wondering if thats the issue or if anyone has any experience with the same bug

Also, I'm trying to save some defaults but certain values like Sampling, Height, and Width (to name a few) aren't saving as defaults

[boreal vector]

I did a clean install and tried Nearest and SwinIR, they didnt work.
I tried LSDR and got this error

i just added --no-half in bat file and now its doing something

[ornate elk]

Yea thats normal for the first image. It can take 15-40 mins for the first time only.

[ornate elk]

Link is broken, here is the direct link:
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides#nvidia-automatic1111-webui-stable-diffusion-webui

[slender sable]

Thank you for that, I made sure the correct arguements were there and says that its loading with the correct arguments. But I am still getting a grey box output.

[ornate elk]

That mostly means your not using a model as model. A model is bigger than 2gb. Everything below isnt a normal model (checkpoint)

Also flux doesn't work if you tried that

[slender sable]

I followed an online guide and added the CompVis/stable-diffusion-v1-4 which is showing up after a fresh install, its 40+gb in size. I am not sure if I need to select something specific from the model list in the top left or change any other settings. The model was added to the models/stable diffusion folder.

[ornate elk]

Yea thats not the right thing

No model is 40gb big

For normal users at least

You can delete the 1.4 stuff

[slender sable]

Oh ok

Is there a base model that I can use? I thought that was what I had.

I obviously have misunderstood what I was I was downloading. I thought that was the model it worked off.

[ornate elk]

Download any other model from Civitai.com for example Dreamshaper v8
https://civitai.com/models/4384/dreamshaper

And put it into models/stable-diffusion

[slender sable]

Will do, thank you so much 🙂

[ornate elk]

No problem!

[slender sable]

Worked perfectly! Thank you

[boreal vector]

well that was a nothing burger

[ornate elk]

Can You share to original output?

So I can check the metadata

[stone fog]

I'm running on a AMD CPU what should I do to bypass this

[ornate elk]

What's your GPU?

[stone fog]

7800 XTX

[boreal vector]

its a 2x upscale

the 4x was like 25mb

[ornate elk]

Then you need to follow my AMD Automatic1111 with ZLUDA Install Guide from here:
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides#amd-automatic1111-with-zluda

[stone fog]

Is this something I can jump in the middle of if I've already installed SD and gotten to the webscreen?

I have auto 1111

[ornate elk]

Which version?

The amdgpu fork or the default?

[stone fog]

I think default

[ornate elk]

Then you need a clean install

[stone fog]

;-;

[ornate elk]

Because the default only works with nvidia

[stone fog]

got it

[ornate elk]

Can you try upscaling with hires fix instead of Extras ?

Set hires steps to 10, set denois to 0.5

Upscale by 2

[boreal vector]

in img to img tab?

[ornate elk]

In txt2img

[stone fog]

Already borked myself

[ornate elk]

You need to run the command in a CMD and not in terminal

And you need Git and Python 3.10.11 64 bit installed

[stone fog]

I have python installed 3.10.11

How do I open up the folder in CMD again, I'm not actually quite suire even after that guide

[ornate elk]

Do you have git installed ?

[stone fog]

Yes

Open GIT Bash?

[ornate elk]

You can use that too yes

[stone fog]

got it

[ornate elk]

Make sure your open it inside the folder you created before

[stone fog]

already on it, already cloning.

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--use-zluda --update-check --skip-ort

call webui.bat``` Correct?

[boreal vector]

the base image was 512x512

so this is 1024x1024

[ornate elk]

Can't see the metadata in this one

[stone fog]

wound up with this error

[ornate elk]

Have you installed hip SDK 6.1 and restarted the PC?

[stone fog]

Doesn’t that come after running the windows batch file?

[ornate elk]

Can you try ESRGAN 4x Anime6b upscaler ?

[brazen chasm]

Can you use stable diffusion if you don't have nvidia?

[ornate elk]

Yes

[ornate elk]

Nope

If you followed my guide you have to install hip SDK 6.1

[brazen chasm]

Can anyone help me with this issue:
Runtime error: Torch is not able to use gpu; add --skip-torch-cuda-test to COMMANDLINE_ARGS variables to disable this check.

[ornate elk]

What's your GPU?

[boreal vector]

that seems to have done something

[ornate elk]

Looks good

[brazen chasm]

Intel(R) HD Graphics 620

[boreal vector]

so I dont have to upscale in extras tab?

[ornate elk]

Also for pony sdxl a resolution around 1024x1024 is needed

Yep you shouldn't upscale in extras

In img2img I can help you to to upscale

[boreal vector]

whats the workflow for it?

[snow breach]

I just bought a 4080 Super to do SD... I hope I'm not a sucker if a built-in 620 can run SD...

[brazen chasm]

Ummm @ornate elk help with my issue pls.🙏

[ornate elk]

That won't work with stable diffusion.
Only on slow CPU mode

[brazen chasm]

Huh!

What do you mean it won't work?

[snow breach]

You need a decent video card in order to use Stable Diffusion, I don't think it will use your integrated card or CPU out of the box. There is a fork of Stable Diffusion WebUI that might help you if you read through it. But... I think it is difficult to get Stable Diffusion to work with a CPU/Intel card.
https://github.com/openvinotoolkit/stable-diffusion-webui/wiki/Installation-on-Intel-Silicon

[ornate elk]

The 620 Intel is an onboard GPU chip.
You can use stable diffusion on CPU mode. But its insane slow

[upper moth]

Got ComfyUI working on my new machine, when I try to upscale images x2 most of the time it works great, but its happened a couple times my screens start turning on and off, then the PC gets stuck and I have to reboot.

I'm guessing my GPU (AMD 7800xt) is getting overwhelmed? Any suggestions?

[zinc arrow]

hey so ive been out of the picture for a while but i remember you were the one that used to help me with stuff

i wanted to ask, what are some good models that can generate "anything" i ask for (not necessarily NSFW)

[ornate elk]

Hey, do you use comfyui with directml or zluda ?

And which workflow do you use

[ornate elk]

Hey, try out UnstableDiffusers Model. That can do pretty much stuff.
https://civitai.com/models/84040/sdxl-unstable-diffusers-yamermix

[ornate elk]

In img2img, set the resolution to 512x512 when using a 1.5 based model.
For SDXL/pony set it to 1024x1024
Then set the denois below 0.3
And at the bottom enable the SD Upscale script and select an upscaler.
Then generate.

[bright burrow]

Guys, is there any code to benchmark image generation inference on different types of hardware?

[boreal vector]

1. Generate an image in txt2img using highres fix (10 steps, 0.5 denoise, upscale 2x)
2. Send the image to img2img (0.3 denoise)
3. Use SD upscale in the script.

is that it

[ornate elk]

Yep

You can also use denois 0.2 if you get artifacts in img2img

There is also a way to upscale with controlnet. But im not at home to share the workflow

[boreal vector]

sampling steps same as text2img?

[ornate elk]

Yep

[boreal vector]

so i dont resize this thing here

but here

what is tile overlap

[ornate elk]

You can leave it at default

Its the pixels between the tiles

[boreal vector]

so more is smooth

[ornate elk]

Yea

[quaint finch]

I need a bit of help again @ornate elk
Im changing to a new GPU and my sd kept using my cpu instead of my gpu to generate images making the process so long

How do i change it to use my gpu instead?

[ornate elk]

Hey, what was your GPU and what's the new GPU?

[quaint finch]

3070- 3070TI

[ornate elk]

Okay. What's in your webui-user.bat?

Also make sure to install the latest Nvidia driver

[quaint finch]

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--xformers --autolaunch --no-half-vae

call webui.bat

[ornate elk]

Okay then delete the venv and relaunch the webui-user.bat.

And if that won't work you have to reinstall the GPU driver and reboot

[quaint finch]

okay, ill try that

thanks a lot, ill come again with the updates

[quaint finch]

still doesnt work

[boreal vector]

1st base 512x512
2nd high res 1024x1024
3rd upscaled in img2img 2048x2048

[ornate elk]

Can you show me the full cmd log?

[ornate elk]

Not bad 👍

[quaint finch]

venv "D:\AI\stable-diffusion-webui\venv\Scripts\Python.exe"
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Version: v1.9.4
Commit hash: feee37d75f1b168768014e4634dcb156ee649c05
Launching Web UI with arguments: --xformers --autolaunch --no-half-vae
[-] ADetailer initialized. version: 24.9.0, num models: 10
ControlNet preprocessor location: D:\AI\stable-diffusion-webui\extensions\sd-webui-controlnet\annotator\downloads
2024-09-18 20:09:19,797 - ControlNet - INFO - ControlNet v1.1.455
Loading weights [c08544e510] from D:\AI\stable-diffusion-webui\models\Stable-diffusion\revEnginePonyxl_v10.safetensors
2024-09-18 20:09:20,223 - ControlNet - INFO - ControlNet UI callback registered.
Running on local URL: (removed)

To create a public link, set share=True in launch().
Creating model from config: D:\AI\stable-diffusion-webui\repositories\generative-models\configs\inference\sd_xl_base.yaml
D:\AI\stable-diffusion-webui\venv\lib\site-packages\huggingface_hub\file_download.py:1132: FutureWarning: resume_download is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use force_download=True.
warnings.warn(
Startup time: 10.2s (prepare environment: 2.2s, import torch: 3.0s, import gradio: 0.7s, setup paths: 0.8s, initialize shared: 0.2s, other imports: 0.4s, load scripts: 1.8s, create ui: 0.6s, gradio launch: 0.3s).

[boreal vector]

i can see the details more clearly now, thanks.

to re enable extensions, do i just put them in the new folder?

or do i have to install them again

[ornate elk]

Install them again

Thats the cleaner option

[ornate elk]

Looks normal.
Can you show the taskmanager performance tab when generating an image?

[quaint finch]

so i did something, i added skip cuda on my arguments and it did get me to use my gpu

but idk why, the process is very long

idk what cause this pyramid thing

it has never happened before

[ornate elk]

Thats very Strange

Have you reinstalled your GPU driver before?

You also need --xformers --medvram-sdxl in your webui-user.bat

[remote cosmos]

What does --medvram-sdxl vs just --medvram?

[ornate elk]

--medvram loads models partially into the vram and helps to not get out of memory when using larger models on low vram GPUs. (6gb and less)
--medvram-sdxl does the same but only for sdxl models. Which improves the generation speed on 8gb vram GPUs

[remote cosmos]

Interesting. Would it work with 6GB cards? My 1660ti takes like 120s+ per iteration on SDXL models (so I don't use them), but would that option allow me to potentially use SDXL, or should I just stick to 1.5?

[ornate elk]

Nope you need --xformers --medvram --no-half
But what improves sdxl speed is when you enable FP8 mode for sdxl in the webui settings

[quiet matrix]

So I just updated forge ui today, and my memory usage skyrocketed, it wasn't like this before, any idea what changed in the settings of the update to cause this? The version I used before was the version before Flux

[ornate elk]

Oh wow. Which settings you use use? Model, and resolution etc

[remote cosmos]

Yep thats exactly what I've been using, and SDXL generation is still taking eons. Is SDXL just not recommended for <8GB?

[ornate elk]

So you tried the fp8 mode already ?

How long does 1 1024x1024 Euler a image takes ?

[quiet matrix]

I've always used these settings with ponyxl, before the update there's no distilled cfg scale

[remote cosmos]

How do I enable fp8 mode?

[remote cosmos]

For 1.5 or SDXL?

For 1.5 its about 3-5s.

For reference I'm on A1111

[ornate elk]

In the webui settings

Search for fp8 and enable it only for sdxl

Then hit apply and relaunch

[remote cosmos]

I got to Settings but dont see any WebUI settings

[ornate elk]

I mean the settings

Your already in the webui

[remote cosmos]

Oh I see specifically "Enable for SDXL" when I search the setting

neat let me try that and redownload an SDXL model lol

I tossed em all out because generating on them was miserable

Is Pony considered SDXL?

[ornate elk]

Yep

[remote cosmos]

Got it. I'll give it a shot, thanks!

[ornate elk]

Thats strange then. What's your GPU ?

How do your settings at the bottom bar look like?

[quiet matrix]

NVIDIA GeForce GTX 1660 Ti

[quiet matrix]

which?

[remote cosmos]

Laptop squad

[ornate elk]

All

[quaint finch]

i got a good solution

set COMMANDLINE_ARGS= --opt-channelslast --xformers --always-batch-cond-uncond

set PYTORCH_CUDA_ALLOC_CONF=garbage_collection_threshold:0.9,max_split_size_mb:512

[quiet matrix]

you mean this?

[quaint finch]

then i added the --medvram-sdxl and it worked well

[remote cosmos]

Crazy, this helped a ton, SDXL now runs about two times slower than SD1.5 on my machine, but that is much better than the 100 times slower it used to be, making SDXL usable for me now!

[quiet matrix]

or this?

[fair oxide]

what is missing in iopaint?? I can't choose any models to erase because there isn't a model

stable diffusion tab is blank with nothing in the list

[ornate elk]

Perfect 🙂

[ornate elk]

This. Make sure you set the mode to xl when using sdxl or pony

[quiet matrix]

still the same

should i use reforge? is that better?

[loud pond]

guys looking to upgrade my gpu what should i get? (i preferable want a gpu with a ton of vram so it can generate images faster)

[quasi ginkgo]

In my browser Inspect console, I'm getting repeated and continuous warnings - "Attempted to select a non-interactive or hidden tab" in the file Tabs.svelte:84:11

My network tab is also doing some GET and POST calls repeatedly

My entire webui has been laggy so I'm trying to troubleshoot it, idk if that could be the reason.

[steady thistle]

I just got a desktop with an Nvidia RTX 4070 Ti Super that was reasonably priced, but make sure you get Nvidia if you care about AI.

[loud pond]

alright where can i buy them from?

cuz nvidia is normally bought out

[steady thistle]

I got it from Best Buy myself, I waited until there was a sale and grabbed one.

I didn't buy the card by itself, it came with the rig.

[loud pond]

oh

alright got it

i need to find a best buy equivalent tho other wise imma be paying fees for shipping xd

cuz best buy is usa right?

[steady thistle]

yes

[loud pond]

alright thanks ill look into it

[ornate elk]

You can try enable the fp8 mode when using sdxl models

[quasi ginkgo]

My forge webui is very laggy. Gen times are fine but it'll take a few seconds to register keystrokes.

[boreal vector]

is there an extension which helps testing settings like this?

[ornate elk]

Its at the bottom of txt2img under scripts, called X\Y\Z

[mint wave]

hey all! I'm trying to get SD to do img2img for human figures in some crappy low-res low-light images. Think trying to put a hat on a person captured in cheap dash cam footage. SD can draw people wearing hats fine, but they're just not very close to where the original people were, even with strength 0.05. My guess is "it doesn't recognize that there's a person there" because the quality is so bad. Any prompting tips for this use case?

[mint wave]

update: fixed a bug, str 0.05 does produce an image much like the original, but it basically just gets blurrier until s=0.25 where random macro scale artifacts start appearing.

[stone fog]

@ornate elk So I'm a little confused on step 4 of your guide.

[boreal vector]

when creating some larger images i get the cude out of memory error, but it goes away if i restart web ui. is there a way i can clear the cached memory?

[ornate elk]

What's your GPU ? And what step isnt clear at 4 ?

[stone fog]

7800 XTX

The pathing

[ornate elk]

Nope. But you can install the tiled VAE extension. That can help to prevent going out of memory

[ornate elk]

Ah the paths in system environment

You need to go here

[dreamy mango]

after adding the zluda to the path, i run the webui-user.bat and get a terminal window saying " couldn't launch python, exit code: 9009 .

[boreal vector]

it added like 3 extensions, I just use the 3rd one?

[dreamy mango]

nvm had to reinstall python from the windows store

[stone fog]

Right Im there but not sure where to go from there

[dreamy mango]

you click on new, and then add the path that it tells you in the guide, something like c:\zluda

[stone fog]

Like this?

[hasty orbit]

If I set hires steps to 0, would it do 20 given how sampling steps is also at 20?

[stone fog]

WARNING: ZLUDA works best with SD.Next. Please consider migrating to SD.Next.
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug  1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Version: v1.10.1-amd-5-gd8b7380b
Commit hash: d8b7380b18d044d2ee38695c58bae3a786689cf3
Using ZLUDA in C:\SD-Zluda\stable-diffusion-webui-amdgpu\.zluda
Installing clip
Installing open_clip
Cloning assets into C:\SD-Zluda\stable-diffusion-webui-amdgpu\repositories\stable-diffusion-webui-assets...
Cloning into 'C:\SD-Zluda\stable-diffusion-webui-amdgpu\repositories\stable-diffusion-webui-assets'...
remote: Enumerating objects: 20, done.
remote: Counting objects: 100% (20/20), done.
remote: Compressing objects: 100% (18/18), done.
remote: Total 20 (delta 0), reused 20 (delta 0), pack-reused 0 (from 0)
Receiving objects: 100% (20/20), 132.70 KiB | 1.18 MiB/s, done.
Cloning Stable Diffusion into C:\SD-Zluda\stable-diffusion-webui-amdgpu\repositories\stable-diffusion-stability-ai...
Cloning into 'C:\SD-Zluda\stable-diffusion-webui-amdgpu\repositories\stable-diffusion-stability-ai'...
remote: Enumerating objects: 580, done.
remote: Total 580 (delta 0), reused 0 (delta 0), pack-reused 580 (from 1)
Receiving objects: 100% (580/580), 73.44 MiB | 9.96 MiB/s, done.
Resolving deltas: 100% (282/282), done.
Cloning Stable Diffusion XL into C:\SD-Zluda\stable-diffusion-webui-amdgpu\repositories\generative-models...
Cloning into 'C:\SD-Zluda\stable-diffusion-webui-amdgpu\repositories\generative-models'...
remote: Enumerating objects: 1064, done.
remote: Counting objects: 100% (123/123), done.
remote: Compressing objects: 100% (75/75), done.
remote: Total 1064 (delta 55), reused 94 (delta 47), pack-reused 941 (from 1)
Receiving objects: 100% (1064/1064), 53.61 MiB | 13.53 MiB/s, done.
Resolving deltas: 100% (545/545), done.
Cloning K-diffusion into C:\SD-Zluda\stable-diffusion-webui-amdgpu\repositories\k-diffusion...
Cloning into 'C:\SD-Zluda\stable-diffusion-webui-amdgpu\repositories\k-diffusion'...
remote: Enumerating objects: 1345, done.
remote: Counting objects: 100% (646/646), done.
remote: Compressing objects: 100% (86/86), done.
remote: Total 1345 (delta 604), reused 561 (delta 560), pack-reused 699 (from 1)
Receiving objects: 100% (1345/1345), 239.07 KiB | 1.93 MiB/s, done.
Resolving deltas: 100% (944/944), done.
Cloning BLIP into C:\SD-Zluda\stable-diffusion-webui-amdgpu\repositories\BLIP...
Cloning into 'C:\SD-Zluda\stable-diffusion-webui-amdgpu\repositories\BLIP'...
remote: Enumerating objects: 277, done.
remote: Counting objects: 100% (183/183), done.
remote: Compressing objects: 100% (46/46), done.
remote: Total 277 (delta 145), reused 137 (delta 137), pack-reused 94 (from 1)
Receiving objects: 100% (277/277), 7.04 MiB | 11.70 MiB/s, done.
Resolving deltas: 100% (152/152), done.
Installing requirements
Skipping onnxruntime installation.
You are up to date with the most recent release.
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
C:\SD-Zluda\stable-diffusion-webui-amdgpu\venv\lib\site-packages\pytorch_lightning\utilities\distributed.py:258: LightningDeprecationWarning: `pytorch_lightning.utilities.distributed.rank_zero_only` has been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it from `pytorch_lightning.utilities` instead.
  rank_zero_deprecation(
Launching Web UI with arguments: --use-zluda --update-check --skip-ort

rocBLAS error: Cannot read C:\Program Files\AMD\ROCm\6.1\bin\/rocblas/library/TensileLibrary.dat: No such file or directory for GPU arch : gfx1036

rocBLAS error: Could not initialize Tensile host:
regex_error(error_backref): The expression contained an invalid back reference.
Press any key to continue . . .``` This is where I get locked up at

[ruby yoke]

The error says it can't find the rocblas/library. Is it there?

[stone fog]

It’s there, not sure what it’s saying

[knotty turret]

can someone help me AnimateDiff not working

[ruby yoke]

it's specifically looking for a file named TensileLibrary.dat file. is that file ther? does it have that name?

[stone fog]

It would seem not

Where does that file come from, why would it not be there?

[ruby yoke]

since you're using zluda - this is probably a question for @ornate elk

[stone fog]

So it would seem

[ornate elk]

It tries to use your CPUs igpu instead of your normal GPU.
So to fix this you need to go into the device manager and disable the Radeon TM Graphics under Display Adapters.
Then relaunch

[stone fog]

Do I have to do this every time?

[ornate elk]

Yes

[ornate elk]

Nope

It stays disabled. But sometimes a driver update could reenable it

[stone fog]

WARNING: ZLUDA works best with SD.Next. Please consider migrating to SD.Next.
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug  1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Version: v1.10.1-amd-5-gd8b7380b
Commit hash: d8b7380b18d044d2ee38695c58bae3a786689cf3
Using ZLUDA in C:\SD-Zluda\stable-diffusion-webui-amdgpu\.zluda
Skipping onnxruntime installation.
--------------------------------------------------------
| You are not up to date with the most recent release. |
| Consider running `git pull` to update.               |
--------------------------------------------------------
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
C:\SD-Zluda\stable-diffusion-webui-amdgpu\venv\lib\site-packages\pytorch_lightning\utilities\distributed.py:258: LightningDeprecationWarning: `pytorch_lightning.utilities.distributed.rank_zero_only` has been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it from `pytorch_lightning.utilities` instead.
  rank_zero_deprecation(
Launching Web UI with arguments: --use-zluda --update-check --skip-ort

[stone fog]

Is there any forseeable problems with it being disabled?

[ornate elk]

Nope. Disabling it can also boost the general performance sometimes

[ornate elk]

Looks good

Brb

[stone fog]

Now I just gota figure out how to get more models

How do I get the anything-V3.0 model?

[ornate elk]

You can download it from Civitai.com

[stone fog]

Do you happen to have a link because all I know how to use are CKPT files

[ornate elk]

Here:
https://civitai.com/models/66/anything-v3

[stone fog]

Can I save settings since some of them have reccomended settings?

[ruby yoke]

Google sheets and just make notes

[stone fog]

If this is a private repository, make sure to pass a token having permission to this repo either by logging in with `huggingface-cli login` or by passing `token=<your_token>````

[ornate elk]

That can be ignored

[ripe nexus]

hh

[azure tiger]

Where are the admins when you need them the most???

SD devs? assistant devs?

[upper moth]

Sorry for the late reply! I was super busy yesterday.
I use a workflow a put together myself with what (I assume) are just the basics, no idea if I'm using directml or zluda, I installed ComfyUI three days ago and am running almost everything from the default, trying not to touch things I dont understand lol

[ornate elk]

Hey no worries ^^ did you followed any guide to install it for AMD, maybe we can see there if you used directml or zluda.
If you downloaded comfyui directly from the official github repo then your using directml

[upper moth]

Yeah the only guide I used was the official instructions, so definitely direct

[knotty sky]

when im trying to start the webui it says command "git" could not be found

what do i do

and the upscaler section doesnt work

[ornate elk]

Yea with directml your limited on the vram and speed

If you want to use comfyui or any other webui with AMD on Windows you can use my install Guides from the pinned messages of this channel

[ornate elk]

You need to have Git installed

[knotty sky]

no i just fixed it by reinstalling git

but i have to test the upscaling because the last time i tried it didnt work

bruh i cant send the error code

    Traceback (most recent call last):
      File "H:\stable-diffusion-webui\modules\esrgan_model.py", line 31, in do_upscale
        model = self.load_model(selected_model)
      File "H:\stable-diffusion-webui\modules\esrgan_model.py", line 49, in load_model
        return modelloader.load_spandrel_model(
      File "H:\stable-diffusion-webui\modules\modelloader.py", line 150, in load_spandrel_model
        model_descriptor = spandrel.ModelLoader(device=device).load_from_file(str(path))
      File "H:\stable-diffusion-webui\venv\lib\site-packages\spandrel\__helpers\loader.py", line 41, in load_from_file
        state_dict = self.load_state_dict_from_file(path)
      File "H:\stable-diffusion-webui\venv\lib\site-packages\spandrel\__helpers\loader.py", line 70, in load_state_dict_from_file
        return canonicalize_state_dict(state_dict)
      File "H:\stable-diffusion-webui\venv\lib\site-packages\spandrel\__helpers\canonicalize.py", line 27, in canonicalize_state_dict
        if unwrap_key in state_dict and isinstance(state_dict[unwrap_key], dict):
    TypeError: argument of type 'NoneType' is not iterable

now

[ornate elk]

Can you show the full log?

[knotty sky]

i just tried to reinstall but it still doesnt work

do you mean that?

[rancid thorn]

venv "D:\AUTOMATIC1111\stable-diffusion-webui\venv\Scripts\Python.exe"
fatal: detected dubious ownership in repository at 'D:/AUTOMATIC1111/stable-diffusion-webui'
'D:/AUTOMATIC1111/stable-diffusion-webui' is on a file system that does not record ownership
To add an exception for this directory, call:

    git config --global --add safe.directory D:/AUTOMATIC1111/stable-diffusion-webui

fatal: detected dubious ownership in repository at 'D:/AUTOMATIC1111/stable-diffusion-webui'
'D:/AUTOMATIC1111/stable-diffusion-webui' is on a file system that does not record ownership
To add an exception for this directory, call:

    git config --global --add safe.directory D:/AUTOMATIC1111/stable-diffusion-webui

Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Version: 1.10.1
Commit hash: <none>
Cloning assets into D:\AUTOMATIC1111\stable-diffusion-webui\repositories\stable-diffusion-webui-assets...
Cloning into 'D:\AUTOMATIC1111\stable-diffusion-webui\repositories\stable-diffusion-webui-assets'...
remote: Enumerating objects: 20, done.
remote: Counting objects: 100% (20/20), done.
remote: Compressing objects: 100% (18/18), done.
remote: Total 20 (delta 0), reused 20 (delta 0), pack-reused 0 (from 0)
Receiving objects: 100% (20/20), 132.70 KiB | 849.00 KiB/s, done.
Traceback (most recent call last):
File "D:\AUTOMATIC1111\stable-diffusion-webui\launch.py", line 48, in <module>
main()
File "D:\AUTOMATIC1111\stable-diffusion-webui\launch.py", line 39, in main
prepare_environment()
File "D:\AUTOMATIC1111\stable-diffusion-webui\modules\launch_utils.py", line 411, in prepare_environment
git_clone(assets_repo, repo_dir('stable-diffusion-webui-assets'), "assets", assets_commit_hash)
File "D:\AUTOMATIC1111\stable-diffusion-webui\modules\launch_utils.py", line 198, in git_clone
run(f'"{git}" -C "{dir}" checkout {commithash}', None, "Couldn't checkout {name}'s hash: {commithash}")
File "D:\AUTOMATIC1111\stable-diffusion-webui\modules\launch_utils.py", line 116, in run
raise RuntimeError("\n".join(error_bits))
RuntimeError: Couldn't checkout {name}'s hash: {commithash}.
Command: "git" -C "D:\AUTOMATIC1111\stable-diffusion-webui\repositories\stable-diffusion-webui-assets" checkout 6f7db241d2f8ba7457bac5ca9753331f0c266917
Error code: 128
stderr: fatal: detected dubious ownership in repository at 'D:/AUTOMATIC1111/stable-diffusion-webui/repositories/stable-diffusion-webui-assets'
'D:/AUTOMATIC1111/stable-diffusion-webui/repositories/stable-diffusion-webui-assets' is on a file system that does not record ownership
To add an exception for this directory, call:

    git config --global --add safe.directory D:/AUTOMATIC1111/stable-diffusion-webui/repositories/stable-diffusion-webui-assets

Press any key to continue

can anyone help out?

[thorny iron]

How to get Clip Skip to show up?

[knotty sky]

i think i might have found the error

[knotty sky]

still doesnt work

now i fixed it

[thorny iron]

Anyone know how to enable the Clip Skip option? :'c

[ornate elk]

Settings, User interface, Quicksettings, there add Clip_show_at_last_layer

Then apply and relaunch

[thorny iron]

Oops I typed 'stop'

It won't let me add "Clip_show_at_last_layer"

@ornate elkDid you mean 'Clip_STOP'? Just to clarify

Also would anyone know why a colour filter suddenly appears on my pictures?

For example.

[ornate elk]

Yes

[ornate elk]

Missing VAE

Add sd_vae in the Quicksettings too and then you have a dropdown

[thorny iron]

Right. God dammit.

Wait a minute. Why was that missing in the first place? Updated version decided to not have it by default? :<

I was missing the sdxl vae, weird.. :<

Some reason when I did an image reverse on prompts from an old output, regenerating it gave me something entirely different.

[rugged latch]

happens

gotta slap stable into submission sometimes

[onyx solar]

Every single setting you have contributes to how an image is generated. A single setting can change the math that determines the vector(s) of inference.

[thorny iron]

Is it because I updated the tool?

[rugged latch]

wut u could so, is for every image u output that u like, copy image into a folder, give folder a meaningful name, copy the params for that photo into folder, and done

[thorny iron]

Hmm.. so I added Clip Skip, maybe if I put it as 0?

[onyx solar]

Tool updates can sometimes change something fundamental that impacts the outcome.

Adding clip skip would absolutely change the math.

[thorny iron]

I changed it to 1, and nothing changed from 2

And I cannot put 0

What's weird is I'm now replicating the source material I originally wanted..

But now I cannot replicate my old pictures.

I guess I 'fixed' something, and miss my mistakes.

Oh oh oh..

When I faff up my tool after doing another reinstall of windows.

Is there a file I can backup / drag over that has all my stable diffusion settings to my new installation?

[ornate elk]

So it worked?

[ornate elk]

Did you moved the stable diffusion folder or did you git cloned it on D ?

[rancid thorn]

no worries i fixed it, thank you tho.

[thorny iron]

If I click this

How do I paste it into A1111?

Or do I have to do it manually

Civitai

[ornate elk]

Paste it into the positive prompt box.
Then click the blue white arrow on the right to auto apply it

[thorny iron]

Thanks!

Upscaling seems to max out my RAM, guess it's time to up to 64GB? ';x

[onyx solar]

It's maxing out your RAM because it's run out of room on the GPU, so it's swapping.

[thorny iron]

Oh shit

Oh well that explains why that problem has stopped happening before where it kept giving me errors.

Since I have 12GB VRAM.

[onyx solar]

That sounds about right.

[thorny iron]

If it's swapping over to my normal RAM, does that mean it's also offloading from GPU doing the workload to my CPU?

[onyx solar]

I only have a 10GB 3080, so I see this happening a lot. But I also have a decently fast 64 GB of system RAM.

No, the GPU is still doing the processing.

[thorny iron]

Oh well then fuck yeah! I'm jamming in 64GB tomorrow! ;d

Whoop

Does this mean I ran out of VRAM, then RAM too?

[onyx solar]

Tough to say that it's not both...the console would possibly give you more info. Either way, you probably just had too big of a job running in some way.

[thorny iron]

I umm..

I THINK I tried to upscale a 1024x1024 up by 2.5?

[onyx solar]

Consider the fact that although 1024² is about 1 megapixel, 2560² is 6.5 megapixels.

[thorny iron]

Wait what.

That's very interesting.

Oh right a Megapixel is literally 1,000,000 pixels.

[onyx solar]

Yes.

So your 2.5 length/width increase is literally a 6.5x jump in total pixels to crunch.

That's a lot.

[thorny iron]

Right. I thought longitude and latitude.

But I forgot to fill in the enormous diagonal gap. ';d

[onyx solar]

Exactly

[thorny iron]

Whoop. x'3 Definitely going to slap in the 64GB and have a lot of fun tomorrow. Hopefully it'll cope.

I DID output a 2x upscale though so, I guess that's my 32GB limit.

[onyx solar]

Same problem is with high-res displays...it's crazy to think that in my lifetime, we go from playing games that were about 57k pixels on the screen at once (NES) to about 8M pixels at once (4k).

[thorny iron]

It's incredible it does so much processing, gigabytes upon gigabytes, fills your RAM out the wazoo.. and what does it pump out in the end?

A 4.25mb png.

[thorny iron]

It is pretty amazing, and of course 8K now being a thing.

Makes me wonder when it'll stop.. I can't help but think 2K is more than enough.. at least on an 18" screen'ed Laptop.

4K on a TV being enough.

8K on a 32" monitor though, I hope I never see one because it might ruin everything for me.

[onyx solar]

It'll never stop. You'll likely find that depth displays will become the next thing. 8K is probably not going to take off at this point, considering some other factors in play. (Note that even now, more people still have HD displays than 4K.)

[thorny iron]

Wait what?

DEPTH..?

[onyx solar]

Layered depth displays is where this stuff goes, in my opinion. There is a lot of work being done to get to that.

[thorny iron]

That's insane.

Honestly I want to say I was born too early, but PS1 and N64 was probably the golden age of gaming. ';d

I hope to learn enough and start pumping out those AI Generated GIFs or Videos.

[ornate elk]

Make sure to set hires steps to 10

And you need xformers in your webui-user.bat

[thorny iron]

But I have a feeling I'd need a proper Desktop setup for that, with a 4090 and 24GB VRam or something.

[thorny iron]

I have that yes.

Hires steps to 10 though let me check what it is now.

It was 0, whoops. ';d

I ASSUME that putting '10' means "Divide picture into 10 segments, upscale one by one and stitch it together" instead of "Do it all in one go, GOGOGO! YOLO!" right?

[thorny iron]

Oh okay, I figured out why my original output wasn't working.

[ornate elk]

Nope.

[thorny iron]

I THINK it's because that PNG to Text doesn't change the model I was using.

The import that is.

[ornate elk]

It means do 10 times a upscaling step. Like a generation step.
0 means, do the same as normal steps, which is to much

[thorny iron]

Oh, despite my wrong assumption, the point is it breaks down the process into smaller chunks so it uses less ram, right?

Or do I still need to put in 64GB ram? ( I don't mind, I want more ram anyway. )

I'm confused now, and very interested to understand more.

[ornate elk]

You don't need more ram. But if you want to upscale sdxl/pony models you need the tiled VAE extension

[thorny iron]

OH...

WAIT..

I had my generation steps at 40

So if my Upscale is 0, it'll try to do 40 upscale steps?

[ornate elk]

Yep

[thorny iron]

Oh my gawd. No wonder. ':f

[ornate elk]

Yea thats to much ^^

[solid night]

I was wondering if there are any extensions that are able to help/assist with prompting ?

[thorny iron]

Yep I fixed it. Turns out that the import wasn't changing my model file. Yaaay it works.

[ornate elk]

Boorutag autocompletion

Can help, its an autocompletion for words. Helps if you don't find the right words

[thorny iron]

I usually find images I like, then reverse them or look at what each model suggests to have in the positive prompt by default, some are a bit crazy like having "score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up," whatever this means.

[solid night]

thank you I'll give it a shot.

[thorny iron]

That sounds interesting.

What's a Wild Card?

Is that just "Insert random value here"? So each generation is not only different, but so are elements like colours of things, or poses etc?

[ornate elk]

Yep

A really good extension for getting different results fast

[thorny iron]

Interesting..

To make my generations change slightly while staying on a seed, I just put in a random letter of the alphabet ';d

[solid night]

I'm running Forge UI on a laptop with both a 3070ti 8GB and a Intel IrisXe IGPU, is it possible to have the image generation use the whole Nvidia GPU and have Windows use the iris?

[thorny iron]

Humm, just a guess, but you might be able to use Nvidia Control Panel with Optimus to restrict what uses what..

By the way when I upscale by 2x, instead of the image being simply upscaled, I'm getting new elements introduced, sometimes even extra limbs.. how do I stop that? :<

Is that step related?

Nevermind I realised there's types of upscalers to use, I think Esrgan x4 is best?

[zinc arrow]

ty ill try it and lyk if its what im looking for

Also wanted to ask what are Pony models

[ornate elk]

Yes, you can open up the nvidia Control panel, there under 3D settings make sure its set to automatic or Not nvidia high performance device.
Then in the webui-user.bat of stable-diffusion add --device-id 1
To make sure it uses the nvidia gpu

[ornate elk]

You define that with the denois slider

Set it to 0.5 or below

[rancid thorn]

can anyone help me out? When trying to generate images all it gives me is blank grey images

[magic lintel]

@ornate elk Can you help me figure out what i'm doing wrong with this lora?

https://civitai.com/models/402667?modelVersionId=448977

https://i.gyazo.com/b694a89e8b37ffcc60485a818d674228.png

[ornate elk]

You need to use a model (2gb or larger) to generate image's.
A lora is not a model and gives grey images

[rancid thorn]

this is what im using and its 15gbs

[ornate elk]

Make a , after green tea, and then use a larger default resolution, because sdxl and pony are trained on 1024x1024. And the pony lora too

[ornate elk]

Flux is currently not supported in auto1111

For that you would need to use Forge Webui or Comfyui

[rancid thorn]

oh

ill look into that and thank you for your help

[ornate elk]

No problem

[magic lintel]

safe to make images like that on a 3070?

[magic lintel]

oh shit

that worked for the slider

thanks

[ornate elk]

Yep if you have --xformers --medvram-sdxl in your webui-user.bat your fine

[steady thistle]

@ornate elk This really isn't a tech question, but could we get a Flux channel? I'm liking that model more than anything else I'd say.

[ruby yoke]

we're sort of using the #🆕｜sd3 channel for flux

[solid night]

Hey thanks for replying earlier, I just got around to trying it out. I'm looking at the console is this normal?

[thorny iron]

You’re the best. Thank you so much!

[ornate elk]

Yea it's normal. You can lower the vram amount with the slider on the top right

[solid night]

Thanks, I only tried a few prompts everything seemed fine. Appreciate the help

[subtle warren]

Umm how can u generate images?

[median cosmos]

Can anyone tell me the optimum hardware to run stable diffusion like which nvidia chip like 3080 or 4090 .Idk if this is the right channel to ask this I

[plain ocean]

is anyone there?

[plain ocean]

nvidia rtx 3090 or 3080 would be a great start

but if you have money go for 4090 without a second look

it really is much faster

[median cosmos]

I have 3080 ti but i am planning on upgrading it

[plain ocean]

then go for 4090 ti

[plain ocean]

dang 3080 ti must already be fast right?

[median cosmos]

Well right now it's on my laptop and I have my vram limited

I am planning on building a PC

[plain ocean]

oh i see

then without a doubt go for 4090 ti

but also do your research too first

and if you are gonna go for the cpu

go for the amds new cpu

idk what the name is

but its far better than having intel

[median cosmos]

Yeah that's what I am doing right now

A lot of PPL in reddit are telling 24 GB VRAM is the best but you need 16 GB at least

[median cosmos]

Will do, personally I have always preferred amd over intel

[plain ocean]

well if you are going for the rtx 4090 then go for the max version itself the 24gb v ram one

it is infact best the higher vram you have mate

and also make sure to build a 32gb ram or 64 gb ram pc

[median cosmos]

Oh okay, but in general is 16 GB the minimum you would atleast need.

[plain ocean]

yeah

goodluck brother

[median cosmos]

Got it mate

And thanks a lot

[plain ocean]

if you need anything else

ask me no problem

[plain ocean]

no problem

i try to help cause no one did to me

[median cosmos]

Alright mate, thanks again

Have a great day

[plain ocean]

you too 🙂

[delicate urchin]

My comfyui has slowed down. Upscale now runs around 30s/it when it used to do 1.2s/it on all 4 tiles when upscaling. See the image. What might have caused this and how do I fix it? Already tried cleanup of vram and the clear cache

[delicate urchin]

restard of pc fixed it 😂

[frigid crypt]

what is the easiest way to update to Forgeui?

[thorn badge]

why does forge keep runnin gout of memory when running a lora in flux?

i have 32GB RAM and a 3080rtx with 12 gb RAM, so should i change these settings?

[odd crypt]

please help

when i first click the bat file and dowloads for a bit

dont know what i am suppose to do

[timber oak]

Hey

im having issues with my comfy

as you can see i have no missing nodes

but my comfy says im missing nodes

nvm i fixed it

thank you reddit

[marsh tide]

what am i doing wrong lol

[gusty compass]

Can anyone help me with a forgui Issue? I can't see any of the loras I downloaded in the lora tab. I put them in the Lora folder, and have refreshed and even restarted forge UI.

[true sphinx]

howdy for some reason textual inversions are not showing up in my UI only some old ones are showing up

[vocal burrow]

you're not trying to use sd1.5 loras with sdxl model ?

or vice versa

[gusty compass]

I'm trying to use a flux model with a flux lora

[vocal burrow]

what does the log says ?

[gusty compass]

sorry what do you mean by log

[vocal burrow]

console log

white text in black window thingy

[gusty compass]

ah

I'm using pinokio, so the terminal is built into it if that's what you mean.

[vocal burrow]

git is not installed, incorrecty python installed (you want anything in between 3.10.6 and 3.10.13)

[vocal burrow]

anything before that ?

[gusty compass]

Microsoft Windows [Version 10.0.19045.4894]
(c) Microsoft Corporation. All rights reserved.

D:\Pinokio\api\stable-diffusion-webui-forge.git\app>conda_hook && conda deactivate && conda deactivate && conda deactivate && conda activate base && webui-user.bat

venv "D:\Pinokio\api\stable-diffusion-webui-forge.git\app\venv\Scripts\Python.exe"
Python 3.10.14 | packaged by conda-forge | (main, Mar 20 2024, 12:40:08) [MSC v.1938 64 bit (AMD64)]
Version: f2.0.1v1.10.1-previous-535-gb20cb4bf
Commit hash: b20cb4bf0e526f890fcd40a4d039da581cfebafa
Launching Web UI with arguments:
Total VRAM 8191 MB, total RAM 32691 MB
pytorch version: 2.3.1+cu121
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA GeForce RTX 3060 Ti : native
Hint: your device supports --cuda-malloc for potential speed improvements.
VAE dtype preferences: [torch.bfloat16, torch.float32] -> torch.bfloat16
CUDA Using Stream: False
Using pytorch cross attention
Using pytorch attention for VAE
ControlNet preprocessor location: D:\Pinokio\api\stable-diffusion-webui-forge.git\app\models\ControlNetPreprocessor
2024-09-20 14:31:23,110 - ControlNet - INFO - ControlNet UI callback registered.
Model selected: {'checkpoint_info': {'filename': 'D:\Pinokio\api\stable-diffusion-webui-forge.git\app\models\Stable-diffusion\flux1-dev-bnb-nf4-v2.safetensors', 'hash': 'f0770152'}, 'additional_modules': ['D:\Pinokio\api
\stable-diffusion-webui-forge.git\app\models\VAE\ae.safetensors', 'D:\Pinokio\api\stable-diffusion-webui-forge.git\app\models\text_encoder\clip_l.safetensors', 'D:\Pinokio\api\stable-diffusion-webui-forge.git\app\m
odels\text_encoder\t5xxl_fp8_e4m3fn.safetensors'], 'unet_storage_dtype': None}
Using online LoRAs in FP16: False

[vocal burrow]

probably a quirk from the bnb nf4 flux model

try a different one if you can

cf https://github.com/lllyasviel/stable-diffusion-webui-forge/issues/1001

[gusty compass]

So loras in forgui will not appear at all, and I just downloaded a 1.5 sdxl lora, dropped it in the lora folder, and loaded checkpoint sd_xl_base_1.0.safetensors

[fierce hollow]

Friends. Forge, flux. Very long time began to generate images - 4 images are generated in the area of 20 minutes (I have a 3070ti) I generate at a resolution of 896x1125

[vocal burrow]

"1.5 sdxl lora," a what ?

[gusty compass]

is what my lora screen looks like

[vocal burrow]

either it is 1.5 or it is sdxl

[gusty compass]

sorry sdxl forget i said 1.5

[vocal burrow]

screeshot your lora folder plz

[vocal burrow]

not sure.... might be a pinokio quirk

try using --lora-dir "lora dir goes here" command line arg

[gusty compass]

would that be from the root folder D:\Pinokio\api\stable-diffusion-webui-forge.git

[vocal burrow]

the whole url

D:\Pinokio\api\stable-diffusion-webui-forge.git\app\models\Lora

[gusty compass]

well I mean where do I type that command? still from that path?

[vocal burrow]

Technically you should put it in the webui-user.bat at the
set COMMANDLINE_ARGS= line

from what I saw in your log it should work

But pinokio has probably some fancy whatever thing in their UI or something to edit the webui-user.bat for you ?

I don't know I never used it

[gusty compass]

I mean I can drop it in the webui-user.bat easy. I'm used to not using pinokio, which has made things easier and more difficult.

[vocal burrow]

well then let's try it

[gusty compass]

still no luck, but maybe it's how I added it.

[marsh tide]

the funny thing is i have 310 installed

[vocal burrow]

what is that ?

[vocal burrow]

yes, 3.10.0

3.10.0 < 3.10.6

[marsh tide]

its now giving em a torch error

[gusty compass]

lol me being a dummy and putting it in the webui and nopt webui-user

[vocal burrow]

ok that's what I thought :p

there should already be a set COMMANDLINE_ARGS=whatever whatever in there

[warm tusk]

With automatic1111, why would some prompts give a NaN error but others do not? For example, simply adding "with lightly tanned skin" to one prompt caused it to start giving NaNs.

[vocal burrow]

Magic of statistics and llm combined with float precision of gpus.

sometimes you're lucky and sometimes you're not

[marsh tide]

See torch error

[vocal burrow]

there are way to mitigate it tho. What's your hardware and command line args ?

[vocal burrow]

delete the venv folder and try again

[warm tusk]

Looks like it is the model, since I often get a purple image with static with some prompts.

[marsh tide]

Same issue

[warm tusk]

4060Ti, AMD 3600XT, --xformers --opt-split-attention --no-half-vae --listen --enable-insecure-extension-access --medvram . And realDreamSDXLClassic_sdxl4.safetensors [eeb912aaab]

[gusty compass]

still no luck. Not sure if there is anything else in here that would be interfering?

[vocal burrow]

python 3.12.6 => too new.
instructions are "any python in between 3.10.6 and 3.10.13"

[vocal burrow]

should be good like that. So that's not the issue here.

[marsh tide]

so its not a pythoin issue its saying its a torch issu3e

[vocal burrow]

yup it cannot find a compatible torch because your python is too new.

[gusty compass]

I might just try forgeui without pinokio, it''s not as convenient, but it's not worth anything if it doesn't work

[marsh tide]

its 3.10 though......

[vocal burrow]

get a compatible python https://www.python.org/downloads/release/python-31011/
install it, delete venv, try again.

[vocal burrow]

this says 3.12.6

if you have multiple python you re probably better uninstalling all of them and simply keeping one that is compatible.

[marsh tide]

im uninstalling the python i have

BTW i had to manualy say where python was ......

well i had to make a folder just for python 310

[vocal burrow]

ok if you know how to point a1111 to a python then you can keep your multiple python and simply install a compatible one and point to that one instead

but it's probably easier to just have only one on your system

[marsh tide]

also

[vocal burrow]

just """for fun""" can you try to tick the "Always show all networks on the Lora page" in your settings->"Extra Networks" tab

[gusty compass]

Yeah I tried that earlier, it was already checked, and I unchecked and rechecked it, but no luck.

[warm tusk]

@vocal burrow Ok, --no-half fixed the error but it just gives a weird image where it is mostly one color with different colored "static" with the same prompt. Is it normal for some models to do that?

[gusty compass]

Thanks for the help regardless

[vocal burrow]

--no-half is probably overkill it's gonna tank your perf and vram usage

what's your model and gpu ?

sorry already answered

mybad

reading it

[warm tusk]

Note that realDreamSDXLClassic_sdxl3.safetensors and the others I have tested works fine without --no-half and on the same prompt.

[marsh tide]

................................

this is bull crap im about to give up already befor i had to wipe my PC this shit was working now it wants to eb hella diffcult

[vocal burrow]

1/ --opt-split-attention is probably not gonna help you much
2/ 99% of the time there is no proper reason to run with --enable-insecure-extension-access
3/ --listen will make your url accessible from all your network, be careful
4/ 4060ti with 16 or 8gb vram ?

[vocal burrow]

full log please

[warm tusk]

16GB vram.

[vocal burrow]

ok you can drop --medvram then

[vocal burrow]

About the purple thingy, it might be due to an incompatible VAE being used (sd1.5 vae with sdxl model). Some rare model can be extra picky and require a very specific VAE.
Regarding the "colored static" output, usually it's because of incorrect settings; cfg too high, outputting too low resolution (512x512) with sdxl model, etc

[marsh tide]

where do i find that log

[warm tusk]

Same settings with v4 as with v3, and the purple with colored static with v4 was only after I tried --no-half. I am tempted to just avoid v4 since every other model I have tried works fine with the same prompt.

[vocal burrow]

that's the thing you're screenshoting. Just either screenshot the whole thing.
Or copy paste it into a .txt and drop said file in here

[vocal burrow]

I already gave you so much stuff to do I'm not sure at that point what you've done and what you haven't tried.
Firstly are those your only command line args now --xformers --no-half-vae --listen ?

--listen you can keep as long as you know what it implies

[hallow latch]

I have problem currently when doing Upscale with Img2Img tab.. I have no prompts what so ever and never had this issue yesterday.

Why is this happening?

[vocal burrow]

you're gonna hate it

[warm tusk]

Yes. Still the same issue with that model.

[vocal burrow]

oh can you screenshot your webpage so I can see what settings you're using to get funky output ?

[vocal burrow]

Also you're installing directly in your users folder, I would not recommend that

you might want to check one of the guide from the pinned message

one thing that might be happening about your python situation is microsoft store

you might have the python from microsoft store installed and it might not show up along the usual "installed programs"

(one more reason to not use microsoft store)

[warm tusk]

Ok, here you go. Removing the "with lightly tanned skin" from the prompt makes it work as expected, more or less.

[marsh tide]

I get the same thing froma diffrent folder as well

this is what redit says to do

[vocal burrow]

not totally the same errors
1/ it's still using python 3.12.6 so either uninstall it or point A111 towards the correct python with the set PYTHON=whateverpath\python.exe from webui-user.bat
2/ before when using users folder you had permission issues because you didn't own the folder. now you have issues because your D: is most likely formatted in a format that "does not record ownership" (most likely fat32 where you should probably be using ntfs instead). Solution for that just run the command it's giving you git config —global —add safe. directory 'D:/AI shit/sd.webui/webui'

[vocal burrow]

don't need it

[marsh tide]

the drive its being run off atm is EXfat

[vocal burrow]

ok that also does not support owernship operations

doesn't really matter

firstly you should fix your python situation.

[marsh tide]

so should i move the folder someplace else?

[vocal burrow]

if you have some free space in c: just dump it in c:\AI for instance

[vocal burrow]

downloading it, giving it a shot, please wait

[vocal burrow]

it's a little heptic atm, you'll have better luck asking in #📝｜prompting-help I think. You might also want to include your settings / screenshot your webpage. I'm seeing a ton of tiling artefacts in there.

[marsh tide]

be done witht he copy soon lol

[vocal burrow]

probably faster to just git clone from scratch

you're probably copy pasting an incorrect venv anyways

[marsh tide]

naw wont take long at all

[marsh tide]

I will just delete that venv

[ornate elk]

A 3080 has only 10gb vram.
But you have set 0gb in forge at the top slider

[ornate elk]

No space left on C drive

[vocal burrow]

Ok I'm blaming the model lol. It's acting real weird with the prompt.
Remove (photorealistic), (realistic) from your prompt and you should get what you want tho.

not screenshoting the result I have for obvious reasons.

sdxl is not fan of the "world salad" prompt method anyways

[warm tusk]

Ok, thanks. I have no idea why it does that but I will keep that in mind if I use that model again.

[ornate elk]

Your on an old webui version. An you should also go over your browser extensions sometime xD

[fierce hollow]

Help me with flux. I installed diffusion_pytorch_model (https://huggingface.co/InstantX/InstantID/tree/main/ControlNetModel).
But it is not displaying for me....
Below I will attach screenshots from Forge, and yes, I updated on the blue button all 3 items, nothing opened... Forge I also restarted ... twice, please help me...
https://imgur.com/a/JMfllKp

[warm tusk]

I am on v1.10.1, commit hash 82a973c04367123ae98bd9abdf80d9eda9b910e2 .

[ornate elk]

Okay, thats the latest. But how did you get the facerestore back there

[hallow latch]

It was the Denoising Strength 😄 It was set on 0.75 I always run it on 0.2 - 0.3 😄

[vocal burrow]

that will do it :p

tldr the higher it is the more it will try to "make sense from the noise"

[warm tusk]

What do you mean? I have been upgrading webui from older versions every so often via git pull and I do run on a TV that is 1360x768 resolution if you are talking about where everything is on the screen.

[vocal burrow]

so if it's too high it will make sense "too quick" and will just reimagine the whole thing

[ornate elk]

You should reinstall stable diffusion, but you need to uninstall Python 3.12 before

[ornate elk]

There was an update that removed the face restore option from txt2img. Thats why I was confused ^^

[marsh tide]

I only have 310 installed

[vocal burrow]

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

[marsh tide]

ok why is it 312

[vocal burrow]

told you, that's the microsoft store's python

@marsh tide

[marsh tide]

this is the only python in the system

[vocal burrow]

it won't show up in there

need to uninstall it from microsoft store

[warm tusk]

Microsoft store apps are handled separately.

[vocal burrow]

they need to kill microsoft store, fast.

(spoiler they won't)

[ornate elk]

Ah thx

[warm tusk]

I am surprised they haven't moved to have all all "legacy win32/win64 software" run in a VM and only allow store apps outside that VM based sandbox, thus no more things like openshell.

[warm tusk]

Yeah, I didn't like having to go into settings every time so I simply put it back.

[vocal burrow]

should be much better now.
delete venv and try again.

And with that I'm off

[warm tusk]

Same here.

[marsh tide]

Das is new

[ornate elk]

Looks good

[marsh tide]

sofar so good

[marsh tide]

so how do i get it to open up in microsoft edge?

[ornate elk]

It always opens in the default browser

[marsh tide]

cause it will auto open up in chrom if i let it

[ornate elk]

You would need to set edge to the default

[marsh tide]

I remeber there was a way to tell it hay open up in this browser instead

whats this?????

[ornate elk]

Have you run the update.bat before the run.bat ? If not this can happen

[marsh tide]

I have no update.bat

[ornate elk]

Hmm strange

[marsh tide]

can you send me that bat file?

these are the files i have

[ornate elk]

Did you downloaded the webui as .zip file ?

[marsh tide]

yes

[ornate elk]

Thats the wrong way to install it

Check the install guide in the pinned messages of this channel how to install it correctly
Here:
#🤝｜tech-support message

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

You need to have Git installed to git clone the repo

Uodsge your extensions too

[marsh tide]

looking better

[ornate elk]

Check the cmd log if there are any errors

[marsh tide]

im just waiting for it to do all its fun shit then i can move my modles over

[ornate elk]

You need to download the insightface. WHL file and then install it inside the venv via pip command

[marsh tide]

it says launching webui with arguments now

[ornate elk]

Thats good. Gives you a better performance

[marsh tide]

WOOT now i can move my modles over and my output folder as well

[ornate elk]

Yep your ready

[marsh tide]

I have a124 gigs of modles and VEA

[ornate elk]

If you have not much space on C you don't need to move the models. You also can link them via --models-dir "D:/PathToModels"

In the webui-user.bat

[marsh tide]

I have 2 TBs on C drive

[ornate elk]

Alright

Why scripts ? But np

Venv

You first have to activate the venv

Then install the .whl file

[marsh tide]

where do i get this lora

[ruby yoke]

i'd do a google search for that and see if a site comes up

[solid night]

Not exactly a technical support question but, Does using a SD that's made for lower VRAM give you worse/different results than using one that's made for lets say 16-24GB of VRAM or is it just faster?

[steady thistle]

Is there a good walkthrough for training one's own LoRA for Flux?

[ruby yoke]

both

[bleak mirage]

Hello I need help figuring out what is going on here 😬

I tried to launch A1111 and this showed up 😬

[sonic scroll]

if anyone can assist me, im having trouble getting my comfy ai manager to work, getting an error, please dm me

[ornate elk]

Nope, all can get the same results.

[solid night]

Out of curiosity what would be the benefit of a full SD vs the same one but pruned?

[ornate elk]

A full model has better precision. But its not worth using it because the file size is doubled

[ornate elk]

Can you show the full cmd log?

[solid night]

Alright thank you.

[ornate elk]

No problem

[mortal compass]

Tried to boot up the new forge after not touching my local AI stuff at all for all summer, but it seems computer says no. I ran the update, and cmd updated it and when I check it again it says it is up to date. I think I have the wrong pytorch and xformers versions? I tried to install this "pip install torch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1 --index-url https://download.pytorch.org/whl/cu121" and it seemed to work, but I'm getting the same error. I also tried reinstalling xformers, writing "--re-install xformers" (or something close to that, can't find it) in ARGS, which did not work.

[ornate elk]

Delete the venv folder and relaunch

[mortal compass]

I was sitting here thinking if I should do that or wait for a reply first Will give it a shot

[mortal compass]

It worked somewhat; at the end the local server started but the interface is not working as intended. There was a whole bunch of errors that I'm not familiar with.
I found it interesting that it installed albumentations and insightface and then decided that it is actually not installed 😂

[snow breach]

I wish I could get SD Forge to work too. Tried their recommended installer and it wont install. Claims my 4080S doesn't support CUDA lol. I beg to differ.
nvcc: NVIDIA (R) Cuda compiler driver Copyright (c) 2005-2023 NVIDIA Corporation Built on Wed_Feb__8_05:53:42_Coordinated_Universal_Time_2023 Cuda compilation tools, release 12.1, V12.1.66 Build cuda_12.1.r12.1/compiler.32415258_0

[bleak mirage]

As in the entire command prompt process? 🤔

[wet fiber]

venv "C:\stable\stable-diffusion-webui-amdgpu\venv\Scripts\Python.exe"
ROCm Toolkit 6.1 was found.
Python 3.10.11 (tags/v3.10.11:7d4cc5a, Apr 5 2023, 00:38:17) [MSC v.1929 64 bit (AMD64)]
Version: v1.10.1-amd-9-g46397d07
Commit hash: 46397d078cff4547eb4bd87adc5c56283e2a8d20
Using ZLUDA in C:\stable\stable-diffusion-webui-amdgpu.zluda
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 2.3.1+cu118)
Python 3.10.11 (you have 3.10.11)
Please reinstall xformers (see https://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
WARNING:xformers:A matching Triton is not available, some optimizations will not be enabled.
Error caught was: No module named 'triton'
C:\stable\stable-diffusion-webui-amdgpu\venv\lib\site-packages\pytorch_lightning\utilities\distributed.py:258: LightningDeprecationWarning: pytorch_lightning.utilities.distributed.rank_zero_only has been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it from pytorch_lightning.utilities instead.
rank_zero_deprecation(
Launching Web UI with arguments: --xformers
ONNX: version=1.19.2 provider=CUDAExecutionProvider, available=['TensorrtExecutionProvider', 'CUDAExecutionProvider', 'CPUExecutionProvider']

rocBLAS error: Cannot read C:\Program Files\AMD\ROCm\6.1\bin/rocblas/library/TensileLibrary.dat: No such file or directory for GPU arch : gfx1036

rocBLAS error: Could not initialize Tensile host:
regex_error(error_backref): The expression contained an invalid back reference.
help)

[ornate elk]

Yes

[ornate elk]

Zluda found your CPUs igpu instead of your normal GPU.
To fix this, go into the device manager and disable Radeon TM Graphics under display adapters.

Then relaunch

Also remove --xformers from the webui-user.bat.
Delete the venv folder and relaunch

Xformers is not AMD compatible rn

[ornate elk]

Make sure you have the latest GPU driver and then checkout the install guides of the link in the pinned messages.

[wet fiber]

Thank you, but the photo takes a very long time, longer than on the CPU, do you know what could be wrong? video card (7800 xt)

Compiling in progress. Please wait...
0%| | 0/20 [00:00<?, ?it/s]Compiling in progress. Please wait...
Compiling in progress. Please wait...
Compiling in progress. Please wait...
Compiling in progress. Please wait...
Compiling in progress. Please wait...

[ornate elk]

Full cmd log pls

[wet fiber]

The problem somehow went away on its own

[ornate elk]

Perfect

[oak jay]

Hi all. Forge was working perfectly this morning then I had to restart my computer and now I'm getting this error message, RuntimeError: Your device does not support the current version of Torch/CUDA! Consider download another version:
https://github.com/lllyasviel/stable-diffusion-webui-forge/releases/tag/latest

But, I installed Forge since the dates shown in the link, so, not sure why it would suddenly not support that version

[ornate elk]

Which of these versions did you installed ?

[oak jay]

How would I find that out?

venv "C:\Users\Philip\stable-diffusion-webui-forge\venv\Scripts\Python.exe"
Python 3.10.11 (tags/v3.10.11:7d4cc5a, Apr 5 2023, 00:38:17) [MSC v.1929 64 bit (AMD64)]
Version: f2.0.1v1.10.1-previous-536-gbc0dece3
Commit hash: bc0dece35671209565309a1c783641d8955f06f8
Traceback (most recent call last):
File "C:\Users\Philip\stable-diffusion-webui-forge\launch.py", line 54, in <module>
main()
File "C:\Users\Philip\stable-diffusion-webui-forge\launch.py", line 42, in main
prepare_environment()
File "C:\Users\Philip\stable-diffusion-webui-forge\modules\launch_utils.py", line 434, in prepare_environment
raise RuntimeError(
RuntimeError: Your device does not support the current version of Torch/CUDA! Consider download another version:
https://github.com/lllyasviel/stable-diffusion-webui-forge/releases/tag/latest
Press any key to continue . .

[ornate elk]

You cant but if it says not supported its the wrong one

Make sure your GPU driver is updated. That can fix it sometimes

[oak jay]

I tried that, it is up to date. Weird that it was working fine before restarting my computer. I

[ornate elk]

Could be lucky that it worked until a restart which reloads drivers

[oak jay]

Ahh, ok. How do I get it working again?

[ornate elk]

You need this version:
https://github.com/lllyasviel/stable-diffusion-webui-forge/releases/download/latest/webui_forge_cu121_torch231.7z

Its cuda 12.1 with torch 2.3.1

Which works good on newer nvidia cards

[oak jay]

If I install that, will it overwrite everything already in the folder, like, my models

[ornate elk]

You can't overwrite

You have to move your models over

Its an seperate instance

[oak jay]

Oh, ok, thanks for you help

So, I'll lose all my extensions too??

And all my settings??

[ornate elk]

You can move them over too

Copy the models, extensions and output folder over.
And also the config.json and ui-config.json

And replace them

[oak jay]

ok, thanks

[magic lintel]

@ornate elk Is there a trick to prompting that helps me to speficically change something on one person in the image and not the other

trying to get two girls to fight and want them to look a specific way

[ornate elk]

Thats doable with the Regional prompter extension

There you can define 2 different persons

Or more

[magic lintel]

oh wow?

okay

[magic lintel]

this?

https://github.com/hako-mikan/sd-webui-regional-prompter?tab=readme-ov-file

[ornate elk]

Yws

[magic lintel]

my man

okay thanks

figures someone would make this

[magic lintel]

do you make images or are you just a tech support angel?

also once i install this thing where are the options?

[ornate elk]

Both

[magic lintel]

show me some of your work friend

[ornate elk]

Under txt2img tab

[ornate elk]

Anime or realism?
Or with regional prompter ?

[magic lintel]

anime or realism is fine

with or withou prompter

stuff your proud of

[ornate elk]

okay sure

@magic lintel here are a few

[oak jay]

Even after the reinstall, I'm getting the same error 😢

[ornate elk]

whats your gpu?

[oak jay]

RTX 4080 laptop

[ornate elk]

ah okay then go into the nvidia control panel. There on the right click on 3D Settings. then on the left select Nvidia High Performance Device in the dropdown. Then save and close. Then relaunch Forge

[magic lintel]

The snake looks amazing, very eden vibes, gyaru Alice also looks good

[ornate elk]

Thanks!

[magic lintel]

How do you prompt to get wide shots?

Like top center

[oak jay]

same result

[ornate elk]

restart the Laptop and rthen delete the venv of forge and the nrelaunch

[ornate elk]

i used wallpaper for that one and sometimes putting closeup oder portrait in negative can help

using "wide shot" is hit or miss

[magic lintel]

Ooo wallpaper, haven't thought of that.

[oak jay]

Nope, same issue

[magic lintel]

@ornate elk Best prompt for aerial view. Also do fisheye lens work?

[ornate elk]

for aerial view maybe birds eye view.
Fish eye works with fish eye lense or fish eye effect. if you go to strong on fish eye you get fish ^^

[atomic zealot]

Hello! I am new to the server, and I think I am missing a role, but I'm not sure which one. In the "want to get started with stable difficusion?" pin in general-chat, it says: "Find the most suitable method for yourself by visiting our #⁠unknown channel!" And of course, clicking on #unknown just tells me that's a channel I don't have access to.

(I selected a number of roles, but I guess not the right ones that would give me access)

My OTHER question was.. I'd heard that sd-webui-addition-networks had been deprecated, and that you can put all your Loras into the toplevel models/Lora. However, after restarting, all my loras are gone from the interface. Nothing in the additional networks tab, nothing in the additional networks dropdown.. I suppose I've gotten used to that sd-webuid-additional-networks interface, such that I'm not sure how to use loras in the top-level loras folder now.

[ornate elk]

hey thats a bug. idk which role could see it.
For you lora issue. Make sure your webui is updated.
Also loras/lycoris etc only show up when a compatible model is selected

[atomic zealot]

I'm assuming 'git pull' in the toplevel should fully update it?

All these models are from civitai, could they all be out of date and incompatible?

[ornate elk]

nope

they should work

but you need to make sure that the loras are made for the same model version

sdxl loras work only with sdxl and pony and not with 1.5 for example

[atomic zealot]

Mmm, in my case, I was experimenting by finding pictures on civitai, seeing what checkpoint/loras they used, and downloading/installing them. That was working fine putting them into sd-webui-addition-networks, for the most part, but I started getting errors, and googling those came up with responses saying sd-webui-addition-networks hasn't gotten updates in some time and we should use the native Loras support. Well so far after moving them to the native loras folder, not a single lora shows up now in the webui.

(Also, haha, it's tough googling because SO much information is out there that is out of date now)

[ornate elk]

Yea and its correct that the extension is outdated and can be removed

On which webui version are you?
Check it at the bottom of the webui

[oak jay]

@ornate elk any other ideas please? Still getting the same error.

[ornate elk]

you can try to use the third version of forge which uses torch 2.1

ah nope 2.3.1 is the recommended one

so what you should try is to maybe reinstall your gpu driver

[atomic zealot]

@ornate elk I'm on v1.10.1! git pull says everything be up to date.

I did remove the venv ... just in case. I've had to do that before to fix weird graphical corruptions

[ornate elk]

okay, can you relaunch and share the cmd log, maybe we can see an error when it tries to load the loras

[atomic zealot]

There's not much to it!

(the additional networks extension is still technically loaded, but there's nothing in there anymore)

[ornate elk]

You have to delete the extension as it could overwrite the normal lora functionality.
Then delete the venv folder again to clear the rest of it.
Then relaunch

I'm off for today

[atomic zealot]

As simple as that! Thanks.

[fair oxide]

@ornate elk so my SD A1111 won't startup without an internet connection is there a way round this like some sort of argument I can use to make it start without trying to download anything so it might start up. The whole reason I have it on my laptop is so I can use it when I don't have internet 😭

[snow breach]

Is this the first time you're running the App? After you run it the first time, you'll be able to run it without an internet connection.

[fair oxide]

This isn't my first time running it

I've ran it for months now but I always get the issue when I'm not online

[snow breach]

Odd. I pulled my wifi from my computer and just fired it up without any itnernet at all, works fine.

[fair oxide]

It runs fine without internet but won't start up without it

[snow breach]

Sounds like you've got a git pull in your startup.

[fair oxide]

I can turn the internet off after it starts up and it runs perfectly

[snow breach]

Do you launch it with webui-user.bat?

[fair oxide]

How do I uhh

Not have that

Yeah

I didn't know there was another way 😭😭

Should I be using a different way to start it up?

[snow breach]

No that is fine, you just need to check whats in it.

[fair oxide]

Wdym

[snow breach]

Open webui-user.bat with notepad

[fair oxide]

Okay

Done

[snow breach]

screenshot

[fair oxide]

Sorry, I'm using discord on my phone. Too lazy to open it on my laptop

[snow breach]

That looks ok.

Do you get any errors in the console?

[fair oxide]

Let me pull it up and screenshot it

Umm

Ignore the music 💀

I'm sending a video instead of a few screenshots 😅

Ignore the background too 😭😭

[snow breach]

Yeah it looks like somewhere in your startup files you've got a command to run updates automatically.

Probably breaking due to that.

[fair oxide]

Where can I yeet it

Like where is it so I can delete it

[snow breach]

Ok I think I am wrong. It looks like you're missing some components to the App and it is trying to fix that problem. Someone else may need to help with this. For the next person who does, here are some cleaned up images @fair oxide is seeing...

[fair oxide]

😭

There I go fucking shit up in ways unheard of again

[snow breach]

Haha. I would assume that if you are on the internet and it does run, it would have fixed it. But I'm still a noob myself, so...

[fair oxide]

Ive been using a1111 for a while I just keep breaking it and I don't understand how because I can't understand what it's yelling at me

[snow breach]

ONe thing I could advise is don't use spaces in any folder names when it comes to these apps. They're touchy.

It looks like you did copy/paste it at one point where it may have added " (2)"

[fair oxide]

Ah that's when I was downloading it but my laptop crashed while downloading it (twice)

So it had files downloaded nit not fully installed and I didn't delete the failed downloads so the next times I downloaded it it came out as (1) (this one failed) and (2)

Two is the one that successfully downloaded

I try not to touch anything filewise in a1111 because I have a habit of breaking it really easily. I only mess with the arguments in notepad 😔

[snow breach]

Good strategy.

Well you could try the cli argument: --skip-version-check

[fair oxide]

Hmmm

Lets see

Testing...

Nope

[snow breach]

Sorry 🙂 You need better help than me haha.

[fair oxide]

😭

It's okay

I wish CS1o was here they always fix it no matter how hard I break it 😔

[ruby yoke]

what are the errors it's giving you?

[fair oxide]

@ruby yoke these

I think I got it???

[ruby yoke]

i can't read that. can you do an actual screen shot, not just take a photo with your phone?

[fair oxide]

I used the command argument --skip-install

[ruby yoke]

why?

[fair oxide]

Because I found an article that said do that

And it worked

[ruby yoke]

okay so - you skipped installing - stuff it apparently now wants?

[fair oxide]

Yeah!

[ruby yoke]

so now you ahve to install the stuff you skipped installing

[fair oxide]

Well

No

Because I've been having this issue for a while but it always works as long as I have internet and now I can use it without internet so

I'm not going to update anything

I'm not going to touch anything

[ruby yoke]

so if it's working, even though it's complaining about missing stuff... i'm confused?

[fair oxide]

It works that's all I need if I try anything else I'll fuck it up 10000000%

[ruby yoke]

okay then - what is it you're needing help with? if it's working without the stuff, apparently you don't need the stuff?

[fair oxide]

Well I WAS needing help with it

[ruby yoke]

those are just warnings anyway

[fair oxide]

I just figured it out

[ruby yoke]

okay 🙂

[ruby yoke]

what you should do is what I always do - break it on purpose, then figure out how to fix it. then, if it breaks in the future you'll know how to fix it OR you'll know you can figure it out

[fair oxide]

Nooooooooo 😭

[ruby yoke]

you can always recover by totaly removing it and starting over

[fair oxide]

You don't understand if I mess up I will NOT be able to fix it

[ruby yoke]

that's how you build your skill set

[fair oxide]

I am DUMB

[ruby yoke]

sure you can. if nothing else, you delete everything and start over

[ruby yoke]

you are not. you're just terrified

[fair oxide]

I don't have the internet for that 😭

[ruby yoke]

but you have a brain and you can make notes when you do have the internet

[fair oxide]

😭

[ruby yoke]

you can do it. I mean, you just DID do it. you figured it out

[fair oxide]

I mean yeah but I got lucky 😔

I've never been good with stuff like this, it's all gibberish to me

I can only follow a reddit tutorial looool

[ruby yoke]

but you're evidently learning - you were able to follow a tutorial. that's a good start. you can do this. one step at a time and breath.

[fair oxide]

Hmmmm yeah, youre right. I used to get confused reading tutorials (I didn't understand a lot of terms used liek what a command line argument was) but now I do and was able to solve it on my own for once

[ruby yoke]

you did good 🙂 just need to have some confidence in yourself. no one starts out able to do this stuff - everyone learns - even @ornate elk

[solemn moon]

my body is so ready for forge flux control nets
FYI

[ruby yoke]

your body is ready? what about your brain?

[fair oxide]

I can get control net to make this specific pose I wnt

It's killing me

[ruby yoke]

are you trying to use flux? or sdxl?

[fair oxide]

Is flux better?

Sdxl

[solemn moon]

flux is better but needs more supporting tools

[ruby yoke]

better is a relative word. SDXL is very good.

[fair oxide]

I have some sdxl models downloaded but it all looks trash no matter what I do

[ruby yoke]

yeah - that's not going to happen

[solemn moon]

i honestly go to 1.5 for anything extra than generating

[ruby yoke]

agreed. 1.5 is excellent for all sorts of things

[fair oxide]

Yeah but it's for a specific character and the best models for him are sdxl

[ruby yoke]

okay and the best controlnets are for sdxl too - but i would be using Comfy - not auto1111, forge, or anything else.

[solemn moon]

post a screen shot of what yre trying

[fair oxide]

I have 1.5 stuff but the character doesn't look as good using his available 1.5 models

[fair oxide]

Wdym like the settings?

[solemn moon]

yea

for yr control nets that dont give the output you want

[solemn moon]

both, mental preparations have been in place.
Been praying to the diffusion gods for a line art daily

[fair oxide]

I cant really screenshot on my laptop rn

Hope you don't mind me using my phone to take pics of the screen

[solemn moon]

so chilled

[fair oxide]

Hope this works

Text is readable at least

[solemn moon]

yea def not listening to control net

[fair oxide]

Ignore the prompt on the side that's unrelated

The images it produces looks like this

(Its sending)

It looks liek this or worse

I've even tried doing it through img2img which gets slightly better results but

It's still bad

Like it's DOG

[solemn moon]

bruh, it looks like the pose was copied though o0\

[fair oxide]

Yeah I used the image as the img2img image

I want the pose as exact as possible 😅

I figured it would yield better results and it DID but it's still not good enough

[solemn moon]

Just deleted forge, back to Auto

[fair oxide]

😭

[snow breach]

What did you do to solve your problem Misery?

[ruby yoke]

actually - just delete forge, auto1111, webui and go to comfy

[snow breach]

I just started with forge last night. Got sick of the performance of A1111. Might have to go try comfy.

[fair oxide]

Has anyone gotten a1111 installed on a non-C drive working? I keep getting "dubious ownership" git errors. Or MUST a1111 be installed on C?

[ruby yoke]

you might start by reading through this https://github.com/AUTOMATIC1111/stable-diffusion-webui/discussions/14130 and then talk to @ornate elk once he's online again

[fair oxide]

Yeah, that's what I'd been trying. But I keep getting "fatal: detected dubious ownership in repository at PATH" errors.

Even when I add safe directory entries on that drive in .gitconfig, errors abound. I can get a1111 running after all those safe entry adds, but then I have to do the same for tmp/extensionfolder and then extensions/extensionname. Very cumbersome/painful and hasn't worked yet.

I'll revisit tomorrow

[snow breach]

I started command prompt as administrator and run the webui-user.bat to get around that.

[ruby yoke]

now now now

[fair oxide]

oh that didn't work either lol

[snow breach]

Well don't feel bad, this is all the flux dev model outputs for me, so not much better here lol

[ruby yoke]

beating up the computer usualy isn't very effective.

[fair oxide]

finally.fricking.found.the.solution.

[fair oxide]

doing this helps fix the "access denied" errors to the drive during extension installation, but does not fix the dubius git errors.

I will post an issue to a1111 to document my solution

I needed: git config --global --add safe.directory "*"

to fix "fatal: detected dubious ownership in repository at PATH" errors.

Do I need xformers?

I get "no module 'xformers'. Processing without..."

[ruby yoke]

depends on what you'll be using - probably not.

[fair oxide]

In what case is it needed/useful?

[ruby yoke]

https://huggingface.co/docs/diffusers/en/optimization/xformers

probably not something you're going to need it for, but start there

[oak jay]

That worked, thank you.

[ornate elk]

If your on nvidia you should use it

[thorn pendant]

i want upscale it without eff its pixel art, how can i do it

[vocal burrow]

Probably because your other drives are formatted in fat32.
either you reformat then in ntfs or you run the command it's telling you to.

[vocal burrow]

It's for performance boost, all you have to do is add --xformers to the set COMMANDLINE_ARGS= line of your webui-user.bat

[solemn atlas]

Hello, I have a question. I tried to move my SD models folder to my SSD. I followed this tutorial but somehow it doesn't copy the folder in my webui folder.
https://www.youtube.com/watch?v=9UU3EuZDDVg It's 2:09

This is what I get

It says: You do not have sufficient permissions to perform this operation.

Maybe someone can help me please

[ornate elk]

You only need to add --models-dir "D:\YourPathToModels" in your webui-user.bat

No need to symlink stuff for auto1111

Also --ckpt-dir works only for models and --lora-dir for loras

[solemn atlas]

I moved the whole models folder to my SSD which includes lora and checkpoints. So I still need to only put this? --models-dir "D:\YourPathToModels

because in that tutorial he created a "copy" folder in that webui folder

like this?

[hybrid kindle]

Hello. Im new to all this stuff, installed stable diffusion for amd following guide, now when i trying to create any image im getting this error
what can i do? i also had "no module 'xformers'. Processing without..." message when installing

[solemn atlas]

this is when i open my webui