🤝｜tech-support

[ornate elk]

A lot has changed

[obtuse hare]

the steps I followed in that guide are the exact same as the video showed

[ornate elk]

But this error is a permission error

[obtuse hare]

either way when I followed the guide I still go the error

the guide in the pin

[ornate elk]

Oh okay

[obtuse hare]

how do i fix the error then?

[ornate elk]

Do you have admin Privileges?

Or a sudo password?

[obtuse hare]

yes

it's my computer

[obtuse hare]

Yes

[ornate elk]

Can you run stuff with sudo on Mac?

I'm not the Mac expert, hes afk

[honest storm]

any idea why the image looks like this?

it looks fine until the very end

so it might be something to do with the vae?

[ornate elk]

What are your txt2img settings?

And what model and VAE are you using?

[ornate elk]

Ahh try lower the lora strength

To 0.5 for example

[honest storm]

thats not it

[ornate elk]

Okay then remove the lora and the negative embedding and try again

If that doesnt help,
I think yesmix is a bit older model so you may need to change the sampler to Euler A

[obtuse hare]

its based on unix so yes you can

[obtuse hare]

ERROR: This script must not be launched as root, aborting...

so that sucks

[ornate elk]

Okay, can you change permission of the folders? Then change that from the .cache and further

[ornate elk]

Looks okay now

[obtuse hare]

which folders?

[ornate elk]

/Users/user/.cache/huggingface/hub/

[red totem]

I'm using an SD upscale and this is my result?

[ornate elk]

In img2img? SD upscale script or ultimate upscaler?

[red totem]

img2img yes, sd upscale script

I'm a noob so I'm sure it's completely my error

[ornate elk]

Set the Denois below 0.3 and the resolution to 512x512

[obtuse hare]

what permissions do i change

[ornate elk]

Everyone

Read & write
For the huggingface/hub folder

[red totem]

Can I use the IP Adapter to turn a digital art piece into a realistic looking photograph?

[ornate elk]

Yes!

Thats possible

[red totem]

Could you point me in the direction to do that?

[ornate elk]

Not today, need to sleep now.
But do you have controlnet already ?

[red totem]

yep doing it all through a1111

[ornate elk]

Then select the IP adapter plus and play around with controlnet weight and the End Step

[red totem]

thanks 🍻

[obtuse hare]

how do i use easynegative

[ornate elk]

Did the install worked now?

[obtuse hare]

yes

works but i want to use easynegative

[ornate elk]

You put the easynegative embedding into the embeddings folder.
Then select the additional networks in txt2img there select the embedding and copy it to the negative prompt

[ornate elk]

I converted anime characters into real persons before with ip-apdater. Thats possible ^^ but idk what you want to try

[obtuse hare]

how do i use anything v5?

[red totem]

I got some digital art pieces

[ornate elk]

You download the model file and put it into the models/stable-diffusion folder.
Then select it in the dropdown top left. Also click the reload button.

[obtuse hare]

I don't see an embedding I just see a safetensors file and a pt file

[ornate elk]

Use the .safetensor file

[obtuse hare]

but where do i get anything v5 if you know

[untold marlin]

I dont seem to have this additional network tab. This mean i have a older version of Auto1111?

[ornate elk]

No it doesn't have a name. Its just a selection in txt2img

Where you can click on lora, embeddings etc

[obtuse hare]

do i use the ckpt file?

[ornate elk]

It got removed I see

[ornate elk]

.ckpt files are unsafe so go for .safetensor

[untold marlin]

I have

[obtuse hare]

wdym by unsafe

[ornate elk]

Yes, you can see the Tab lora and textual inversion at the top

[obtuse hare]

removed?

[ornate elk]

.ckpt files are older model files that can contain malicious stuff if somebody upload them.
Like if you see Anything V5.ckpt its a fake because Anything V5 was released as .safetensor file

To stay on the safe side use only .safetensor and your fine

[ornate elk]

Nope it wasn't removed. I found it here:
https://civitai.com/models/9409?modelVersionId=30163

[red totem]

I want to use the Ultimate SD Upscale on a 1024x1024 image to x2 it to 2048x2048 but it keeps saying: ValueError: images do not match

[ornate elk]

Why not using SD upscale script?

[red totem]

it kept just producing blocks

[ornate elk]

Did you changed the Denois value like I said ?

[red totem]

yep

it was a 1024 image though so idk if that does anything

[ornate elk]

That shouldn't produce blocks

[red totem]

blocks being this:

[ornate elk]

Can you show me the whole img2img settings?

[red totem]

yep yep one sec

[ornate elk]

Lol

[red totem]

go easy on me I don't know shit 😦

[ornate elk]

You shouldnt use a refiner or tiled diffusion when upscaling

Tiled Diffusion is the cause for the blocks in this case

[obtuse hare]

I don't see anything that says additional networks

[ornate elk]

Okay then just look for the Textual Inversion tab

Thats the other word for Embedding

[red totem]

so, why would I want to use the Upscale scripts vs. tile diffusion etc

[ornate elk]

Also dont use cfg 25

7-12 is enough

[obtuse hare]

m3 pro macs are crazy

images are taking like 20 seconds to generate

[ornate elk]

Because the upscale script is tiling the image, like tiled diffusion would too. So you activated two tiling methods at once. That caused the issue.
SD upscale script uses less vram
But for upscaling in general you should also use hires fix first in txt2img.

[ornate elk]

Thats epic

M3 is 🚀

[obtuse hare]

I have gen alpha attention span it must generate in 0.018 nano seconds

(joke)

[red totem]

Is there a reason I would use Tile Diffusion over the upscaler though?

[ornate elk]

Nope only if you want to use it together with the controlnet tile model. These two together can upscale. But I didn't liked the result against normal upscaling

@red totem my workflow for upscaling:
For the first upscale use Hires fix.
For the second and third upscale use the SD upscale script.
For the forth upscale use the Extras tab

Then you can get up to 8-16k resolution

[red totem]

dang!

is there any benefit to changing the Latent til width/height for Tile Diffusion?

I'm noticing for what i'm doing (digital art) the upscale scripts make the end result softer and smoother. Tile Diffusion seems to make it sharper and give it more detail

[obtuse hare]

is there a lora that fixes fingers

[ornate elk]

At the top In img2img set the resize mode to "Just resize (latent upscale)"

That will get you more detailes too then

But you may need to remove person specific prompts before

I can't say much about tiled diffusions latent stuff as I only use its regional prompt control function.

I'm now off gn

[sonic ferry]

How do you deal with a long list of LoRA? Do you search it up on Civitai to recheck how to use it? Or is there some way to know this from ComfyUI? :/

[midnight sky]

Colab .. anyone know why dreambooth dose not work ?

[dense lily]

i just upgraded to a 7900 xtx and i'd like to run sd with it, but i'm struggling with getting it working at reasonable speeds. what are my options atm for windows 11 + amd gpu? i think rocm only working on linux atm but don't want to do that and the stable-diffusion-webui-directml i've tried is taking like 20 minutes for generation compared to my previous 3060ti that would be maximum 3 minutes on the same kind of prompts

[hybrid abyss]

Anyone else using Linux? When running webui.sh python packages install into ~/.local/bin. I don't want them there, as having them there previously caused the build of a different app to fail. Anyone know why they are installing there and how to prevent it?

[rotund lava]

try to use comfyui on windows,its faster than webui or go with linux,those are your only 2 options when u have amd gpu

[dense lily]

does comfyui have something that works with amd gpu or is it still directml?

[sonic ferry]

Windows CPU mode user here, is CPU mode on Linux faster?

[rotund lava]

directml

[lavish meadow]

what sampler is the best?

[dense lily]

thanks, thought i got webui to use directml but wasnt, but got directml working on comfyui and will play around with that. thanks

[rotund lava]

yes on comfy directml mode is a little bit faster but if u want better speeds you would have to switch to linux

[dense lily]

yeah thats alright, its usable for now but thought has crossed mind

[lavish meadow]

What is the highest quality sampler?

[stiff cobalt]

for anyone using comfyui in here. ive been trying to copy a generation but i have a question in the Ksampler node about the seed. is that the seed of the image result or something else? the seed of the original image is 3410854465

[stiff cobalt]

also how can i fix this?

[onyx solar]

The seed you see in the box after you've generated an image is the seed that will be used in the next image. If you're trying to recreate an image, you would put the seed value into that box. However, there are many variables in recreating an image you find elsewhere. Even if you have most of the information, you still may not recreate the exact same image for many possible reasons.

[stiff cobalt]

ahh ok ty. and yea its been tricky. ive tried switching to comfy but theres a lot to learn

[red totem]

anytime I try and run a Controlnet type i get this error: RuntimeError: The size of tensor a (96) must match the size of tensor b (256) at non-singleton dimension 3

my image is 1024x1024 and I'm not sure what's wrong

[sonic ferry]

Trying to figure out Ultimate SD Upscale when using LoRA stack, do I link it straight to the Load Checkpoint Model? Cause I thought surely I should link it after the LoRA stack but I can't connect it there.

[jaunty sphinx]

how do i fix prompt size in python script

i ValueError: prompt_embeds and negative_prompt_embeds must have the same shape when passed directly, but got: prompt_embeds torch.Size([1, 154, 768]) != negative_prompt_embeds torch.Size([1, 308, 768]).

[ornate elk]

Hey, I have the same GPU, follow my install guide for AMd Cards, its in the Pinned Messages of this Channel. Using Auto1111 Directml getting 3it/s.

[simple tartan]

oh, dude, you're still online!

[ornate elk]

You need to check if the Model is made for 1.5 or sdxl

[ornate elk]

Still? I just woke up rn xD

[simple tartan]

so I'm pisssed at the ubuntu installer, and i'm going to buld my arch drive 😉 got any tips for me? going to do the A1111 and hopefully comfyUI

[simple tartan]

oh, nice, I had assumed you were US, heh

[ornate elk]

Check this guide for arch:
https://civitai.com/articles/1503/how-to-install-rocm-on-any-form-of-archlinux-or-archlinux-variant

[simple tartan]

awesome soz! thank you mate!

[ornate elk]

Np, Which distro do you gonna use?

[simple tartan]

Arch main

[ornate elk]

Ahh okay

[ornate elk]

I like Manjaro, EndavourOS and Garuda when using arch

And SteamOS!

[red totem]

Trying to inpaint in Img2img and get this error** RuntimeError: "log_vml_cpu" not implemented for 'Half'**

[karmic crown]

in most cases this means that you need to add --no-half to your command line params
what GPU do you have?

[red totem]

I restarted my A1111 and it worked

[acoustic hatch]

hello

in stable diffusion version 2.1 there are unusual drawings in the images. It draws the images worse than normal. i have provided examples below. is it possible for you to check?

Stable diffusion v2.1

[karmic crown]

what dimensions you are using?

[karmic crown]

image dimmensions like 768x768

[acoustic hatch]

ha no 1024x1024

[karmic crown]

and please lets continue this conversation in https://discord.com/channels/1002292111942635562/1011743094309396631 that is right place for those kind of questions

this one is for techical problems regarding installtion and errors

[acoustic hatch]

ha okey

[steady thistle]

Hi all! I just found this discord and I'm struggling hard if anyone can help with AMD Radeon install of Automatic1111. I had it installed and running. It wasn't perfect, but fool that I am, trying to make it better, I ran git pull and everything broke. I tried multiple reinstalls the same way I installed the first time, but I can't get A1111 running and Comfy and Invoke are either incredibly slow or not working either. Processor AMD Ryzen 9 5980HX with Radeon Graphics, 3301 Mhz, 8 Core(s), 16 Logical Processor(s), Radeon RX 6800M GPU with 12GB GDDR6

Can anyone recommend a good set of install directions for my laptop? I'll try yet another fresh install but I'm missing something somewhere and it's been a few days since my last attempt.

I just found the pinned messages... maybe the key is there, I'll try but it's been rough so far.

[lost crescent]

WAS Node Suite - impossible to load in ComfyUI??!?!?!? Why so because?

[lost crescent]

https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Install-and-Run-on-AMD-GPUs

[karmic crown]

what log shows? i do not have expirence with comfy, but i can see if there is anything in log (i am a python developer)

[steady thistle]

Thanks, I'm trying the pinned message right now, but if that fails I'll try this one. I think I tried that last. Pinned showed me 'pip cache purge' which seems like a useful thing. Crossing fingers.

If I can't get this working alright today I'm probably buying an Nvidia. This laptop has 12gb on the GPU so I'm really wanting this to work as it should at the very least be pretty good at AI image gen.

[ornate elk]

hey, use my pinned guide for amd and youll be fine 🙂

[steady thistle]

It's downloading now! So many downloads. So many gigs. But it's coming.

I remember when 1.5TB storage was a lot.

[steady thistle]

Sorcery! It actually worked! I think it was the pip cache purge that did it. I had tried those instructions previously. Thank you so much good person!

[karmic crown]

than you are young 🙂
I started with 48k 🙂

[ornate elk]

perfect 🙂 for any questions about AMDs SD performance and limitations you can ask here, you should get the Tiled Diffusion Extension, it will be needed

[steady thistle]

Oh, no, I was just thinking more recently. I remember a Wang pc with 8" floppy disk.

[steady thistle]

Do tell! Where do I find this? I now trust you implicitly. lol

[karmic crown]

go to extensions tab, than available

type tiled vae

[steady thistle]

I wanted to avoid the obvious Wang joke, but we had one in the house.

[karmic crown]

and install

[ornate elk]

click on Extenstions, Available, Load from, then search for tiled diffusion, click install, then restart

[umbral compass]

Not sure where this question should go, but I'm debating on if I should try ComfyUI out. I saw that Auto1111 has an extension for embedding the ComfyUI into its Tab. or should I install ComfyUI separately as a standalone and link the folder into Auto1111 if I have Lora that I want to use.

Also, is there a difference between install ComfyUI (Directly 7-Zip File) or Manually?

[lost crescent]

[2024-01-05 14:22] 5.0 seconds (IMPORT FAILED): B:\ComfyUI\custom_nodes\was-node-suite-comfyui
[2024-01-05 14:22] 6.2 seconds: B:\ComfyUI\custom_nodes\ComfyUI-Impact-Pack
[2024-01-05 14:22] 8.2 seconds: B:\ComfyUI\custom_nodes\comfyui_segment_anything
[2024-01-05 14:22] 20.3 seconds: B:\ComfyUI\custom_nodes\ComfyUI-LCM
[2024-01-05 14:22]
[2024-01-05 14:22] https_key OK: B:\ComfyUI\custom_nodes\comfyui-mixlab-nodes\https\certificate.crt B:\ComfyUI\custom_nodes\comfyui-mixlab-nodes\https\private.key
[2024-01-05 14:22] Starting server
[2024-01-05 14:22]
[2024-01-05 14:22] To see the GUI go to: http://127.0.0.1:8188
[2024-01-05 14:22] To see the GUI go to: https://127.0.0.1:8189
[2024-01-05 14:22] FETCH DATA from: B:\ComfyUI\custom_nodes\ComfyUI-Manager\extension-node-map.json
[2024-01-05 14:22] data_path: B:\ComfyUI\custom_nodes\comfyui-mixlab-nodes\data
[2024-01-05 14:22] #read_workflow_json_files_all B:\ComfyUI\custom_nodes\comfyui-mixlab-nodes\app
[2024-01-05 14:28] FETCH DATA from: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json
[2024-01-05 14:30] FETCH DATA from: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/extension-node-map.json
[2024-01-05 14:30] FETCH DATA from: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/extension-node-map.json
[2024-01-05 15:01] FETCH DATA from: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/custom-node-list.json
[2024-01-05 15:06] FETCH DATA from: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/extension-node-map.json
[2024-01-05 15:06] FETCH DATA from: https://raw.githubusercontent.com/ltdrdata/ComfyUI-Manager/main/extension-node-map.json

[ivory bramble]

Is there a reason why my final result goes to this?

[steady thistle]

Sounds great. I want to do face swapping and ControlNet depth maps a lot, along with realistic images. Any recommendations on that?

[lost crescent]

Bad sampler

[zealous jackal]

Am I allowed to ask questions about A1111 here?

[lost crescent]

Try a different one

[karmic crown]

[2024-01-05 14:22] 5.0 seconds (IMPORT FAILED): B:\ComfyUI\custom_nodes\was-node-suite-comfyui

I hoped there will be some python error with cause 😦

[ivory bramble]

Ohhh, thank you!

[lost crescent]

Y E S

[ivory bramble]

What is one you recommend for realistic portraits?

https://i.imgur.com/HdxudG9.png

[lost crescent]

Best are DDIM, Euler, dpmm++ 2m

[ivory bramble]

I have these. Sorry about the brightness. My ShareX is brightening my screen captures

Ok, cool! I'll try those out

[lost crescent]

Euler Ancestral, Heun, Uni_PC

[ornate elk]

best would be to install comfyui as standalone by downloading its zip, After downloading and extracting it you need to find the the file called: extra_model_paths.yaml.example, Rename this file to extra_model_paths.yaml and edit it with your favorite text editor. Then at the first line set your path to Auto1111 webui folder. then save. Comfyui now has access to all models from auto1111

[zealous jackal]

Great
So I have python 3.10.6 installed but I get this error:

[ivory bramble]

When I download a new sampler, what folder does the file go in? Sorry, I'm super new to this whole thing.

[lost crescent]

Type where python and find where it is loaded

[ornate elk]

did you installed it while checking "add python to path" ?

[zealous jackal]

Yep

[ivory bramble]

Also, how do I get txt2img to generate piercings? I have tried a million prompts

[lost crescent]

In A1111 settings, the samplers are listed. Check the ones u need to use

[zealous jackal]

C:\Users\ {MYUser}\AppData\Local\Programs\Python\Python310

[ivory bramble]

Oh ok, so you generally don't download new ones?

[lost crescent]

OK, are you using Local or Global SD

[lost crescent]

No, just check the boxes in the list and they'll show up

[ornate elk]

yes, you need to install controlnet and the 700mb models for it, face swapping can be done via the Extension called Reactor or by using the IP-Adapter of Controlnet. Realistic images can be done via realistic models and 4x-UltraSharp upscaler, aswell as the 84000-mse vae

[zealous jackal]

Me? If yes then what's that?

[lost crescent]

U are using A1111?

[zealous jackal]

Yes

[ivory bramble]

Tysm

[lost crescent]

OK, then mebbe u need to reset your PATH via Environment Variables (I'm assuming u use Windows?)

[zealous jackal]

Correct

[zealous jackal]

via Environment Variables
Yeah.. 😬 Where might that be?

[lost crescent]

https://www.java.com/en/download/help/path.html

Look at the Windows 10 section

Click the PYTHON Path - or if its not there - Click NEW

(I know this is for Java, but u can use the same procedure for PYTHON)

[ivory bramble]

Is there an extension that allows you to automatically clean up prompts with the press of a button?

[lost crescent]

Clean up? U mean totally delete?

[ivory bramble]

Lol, no put it into a cleaner format

like a, b, c because I copy a lot of things from chatGPT

and ChatGPT overexplains

[zealous jackal]

Am I tripping?

[lost crescent]

I use a chat GPT Mixlab Node - it really makes your prompts graceful and sophisticated

Click NEW - and then add the python folder address as the new PATH

[zealous jackal]

Ah

alr

[lost crescent]

Of course, call it PYTHON

[zealous jackal]

Top or Bottom new?

[ivory bramble]

Do I need ChatGPT+ for that?

[lost crescent]

TOP

[zealous jackal]

Right done
What about the Value?

[steady thistle]

What's the best way of installing ControlNet with pose and depth map and how do I find the models? I'm sorry for asking so much, I just got gun shy after a week of fighting this.

[ivory bramble]

Also props to you for answering all these questions at once

[lost crescent]

ChatGPT links to nodes via Playground - u pay a small amount and it generates prompts every time u load the Mixlab Node

https://github.com/shadowcz007/comfyui-mixlab-nodes

[ivory bramble]

Awesome, I'll check this

out

[lost crescent]

I spend a lot of time getting the answers which I cannot get here at Discord.

Today I am stuck because Was Node Suite has stopped working.

I am waiting on an answer ... zzz 🙂

[zealous jackal]

?

[lost crescent]

WAS Nodes are now unsupported ... and that might be the only answer?

[hearty igloo]

Downloaded a model and faced with this error during processing.What should i do ? AssertionError: We do not support vanilla attention in 1.13.1+cu117 anymore, as it is too expensive. Please install xformers via e.g. 'pip install xformers==0.0.16'

Failed to create model quickly; will retry using slow method.

[karmic crown]

that looks outdated

what version of a1111 do you have?

check at the bottom of the screen

[zealous jackal]

Guess bro doesn't want to help no more 💀
I changed the value to the python 3.10.6 version but it still says the same

[hearty igloo]

v1.7.0

[karmic crown]

@hearty igloo show us this

[ivory bramble]

When you're inpainting, how do you change the setting that lets you blend better for a more realistic result?

[karmic crown]

it looks like you have new a1111, but old fines in venv folder

[lost crescent]

U can use jan.ai as a FREE alternative to ChatGPT - supposedly quite good

https://github.com/janhq/jan?tab=readme-ov-file

[ornate elk]

you install the controlnet extension like you did with tiled dissusion, but search for sd-webui-controlnet

then download the models (multiple 700mb files) +their config files from here:
https://civitai.com/models/38784/controlnet-11-models
Put everything in the models/ControlNet folder and restart

[steady thistle]

This is all working. You have no idea how happy I am right now.

[hearty igloo]

what can i do

[karmic crown]

please screenshot the whole line as i did, and show me, so i can see all the versions

[ornate elk]

😄 yea for AMD gpus are multiple ways and webuis but on windows its a bit tricky

[ivory bramble]

what's the difference between inpaint and sketch?

[karmic crown]

or from https://huggingface.co/lllyasviel/ControlNet-v1-1/tree/main

[ornate elk]

pls not, they are uneccessary large

[karmic crown]

what GPU do you have? @ornate elk which version AMD use 2.0.1?

[karmic crown]

so, now i need to download new ones, since mine are 1.4GB and not 700MB 😉

[ornate elk]

it makes an impact if your on AMD, as smaller size loads faster, and safes vram

[karmic crown]

you should left me in ignorance 🤣

i would be happy with my 1.4GB models 🤣

[ornate elk]

whats your GPU?

[ornate elk]

in sketch you can draw stuff SD will then try make it to your prompt and transform it,
inpaint will replace something

[hearty igloo]

NVIDIA Geforce GTX 1060 6GB

[ornate elk]

okay, then you need to edit the webui-user.bat (by right click, edit)
at the line COMMANDLINE_ARGS=
you add: --xformers --medvram --no-half-vae
Then save the file.
Then delete the venv folder.
Then relaunch the webui-user.bat
Then go update your extensions in the extensions tab and restart again

[graceful girder]

hi, well let me explain my problem to you in order to find help I use an AMD card and since the last update that I did on my stable diffusion it seems that it is not I don't really want to install the latest version in the Torch list as well as all that goes with it. Could someone help me fix this problem ?

[karmic crown]

@ornate elk this looks like good files, but bad venv to me

[ornate elk]

you installed python 32bit, but you need python 64bit, also you installed the webui in a OneDrive synced Folder, that can cause issues.
Better is to uninstall python and follow my Guide for a Fresh installation from here:
#🤝｜tech-support message

[karmic crown]

ooops

you are right

[karmic crown]

i havent notice 32bit, sorry, CS1o is right

[ivory bramble]

Isn't there an erase brush without having to erase your entire mask?

[ornate elk]

no problem, you cant imagine how long i riddled when this error first appeared 🤣

[hearty igloo]

where is venv folder

[ornate elk]

inside the stable-diffusion-webui folder

[graceful girder]

okay i try this thanks

[karmic crown]

my m1 is 13" and i am tired, it was to small to read properly 🤣 i usually open picture in another window so i can zoom in, but i didnt this time 🤷🏻‍♂️

[hearty igloo]

there is no folder called venv. cant see

[ornate elk]

it should be there, can you make a screenshot?

[kindred tinsel]

Anyone know if there might be an issue if my card says it only supports up to windows 10 and I’m using everything including web ui in windows 11

Would that affect performeance in some way?

[graceful girder]

okay so now i install all thing my stable diffusion need but why torch say me that ? lmao

[broken yew]

do --use-directml and --onnx on the cmdline arguments

[graceful girder]

ok I try that

[ornate elk]

No don't add --onnx

[ornate elk]

Thats because you didn't looked at my guide

You need --use-directml and other commandline args

[graceful girder]

and what is the other commandeline args ?

^^'

ok i go look ur guide

[hearty igloo]

there is no venv folder what should i do

[ornate elk]

you installed the webui the wrong way by downloading it as zip
go one folder back to sd.webui, there should be the venv to delete

[broken yew]

skill issue

[graceful girder]

ok so i folow the instruction but i thinks thats same no ?

[karmic crown]

what is your gpu, cant find that

[graceful girder]

i m on amd gpu

[hearty igloo]

do you mean enviroment

[elder turtle]

@hearty igloo I mean this

[hearty igloo]

there is no venv

oh

[hearty igloo]

--xformers --medvram --no-half-vae

it was clear when i got this error but now wrote those cs1o told me

[karmic crown]

may you please rename folder stable-diffusion-webui-directml to anything else (just as a backup)
and do step 3, but do not close the command line

let me know when you do it

[ornate elk]

show your webui-user.bat content

[karmic crown]

ok, you can do that first

[ornate elk]

it seems like you didnt editet the file correctly

[karmic crown]

that is not good

[graceful girder]

why ?

[karmic crown]

as CS1o wrote in guide:

Edit the webui-user.bat (right click), At the line COMMANDLINE_ARGS= You add: --use-directml --medvram --opt-sub-quad-attention --opt-split-attention-v1 --no-half-vae --upcast-sampling

you are on AMD

[ornate elk]

thats the AMD Guide:
#🤝｜tech-support message

[karmic crown]

thats why i told him to rename his current folder and to do clone and not to close cmd

beacuse i want to see what he cloned first

[graceful girder]

lmao i folow the normal guide

[ornate elk]

Show the inside of SD.webui folder

[karmic crown]

for me, normal guide is the one i wrote (i am on mac) 🙈

[ornate elk]

Okay what's in system?

[graceful girder]

ahhhh my ia can gen correctly and faster now thanks

[ornate elk]

Perfect 🙂 np

[ornate elk]

Okay no clue then, can you use the search function for venv ? If not found you would need a reinstall

[hearty igloo]

how should i reinstall it

[ornate elk]

You can follow my guide for nvidia GPUs here:
#🤝｜tech-support message

[hearty igloo]

File "C:\Yeni klasör\sd.webui\system\python\lib\site-packages\xformers\ops\fmha\dispatch.py", line 69, in _run_priority_list
raise NotImplementedError(msg)
NotImplementedError: No operator found for memory_efficient_attention_forward with inputs:
query : shape=(2, 1024, 10, 64) (torch.float16)
key : shape=(2, 1024, 10, 64) (torch.float16)
value : shape=(2, 1024, 10, 64) (torch.float16)
attn_bias : <class 'NoneType'>
p : 0.0
flshattF is not supported because:
xFormers wasn't build with CUDA support
Operator wasn't built - see python -m xformers.info for more info
requires a GPU with compute capability > 7.5
tritonflashattF is not supported because:
xFormers wasn't build with CUDA support
Operator wasn't built - see python -m xformers.info for more info
triton is not available
requires A100 GPU
cutlassF is not supported because:
xFormers wasn't build with CUDA support
Operator wasn't built - see python -m xformers.info for more info
smallkF is not supported because:
xFormers wasn't build with CUDA support
dtype=torch.float16 (supported: {torch.float32})
max(query.shape[-1] != value.shape[-1]) > 32
Operator wasn't built - see python -m xformers.info for more info
unsupported embed per head: 64

i am able to generate it now but gives this error now.I need reinstall right ?

[ornate elk]

Yes and make sure your GPU driver is updated too

You can copy your models over after the new install

[hearty igloo]

how can i delete all folders?Is it enough to just delete sd.webui folder ?

[ornate elk]

Yes

Its enough

[simple tartan]

probably not for here, but does anyone have a Pocket AI? I'm curious if i should grab one for my LLM stuff, at least until we get rocm in winblows

[lost crescent]

Anybody else has a WAS Node Won't Work Problem at all?

[hearty igloo]

didint understand add python to path how can i do that

[ornate elk]

When you install python 3.10.11

There is a checkbox

"Add python to path"

[hearty igloo]

git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git i will paste this here right ?

[ornate elk]

Yes 👍

But do that in a cmd

Not powershell

But you can try in powershell too

[karmic crown]

is that really matter?

[hearty igloo]

how can i open cmd

using win 11

[karmic crown]

in search type cmd and press enter

[ornate elk]

No then he isn't in the right folder (I know he could CD)

[hearty igloo]

yea

[ornate elk]

You can right click in the folder and select Git Bash here

[karmic crown]

i guess i am too old, in my time anyone new knew how to use cmd and cd

[fading herald]

Hey, any ideas why the cmd would close by itself?

I can't find any sort of crash logs

[simple tartan]

i'm with you there, except now i constantly mix up ifconfig with ipconfig, dir and ls, etc

[ornate elk]

What's your GPU and what's inside your webui-user.bat?

[hearty igloo]

there is no git bash

[ornate elk]

Then try powershell,
should work too

[fading herald]

Gigabyte 4070 Ti Gaming OC and for the webui-user I just added these arguments --xformers --no-half-vae --api --update-all-extensions and a git pull command before the call

[ornate elk]

Okay, what's your python version?

[fading herald]

3.10.11

[karmic crown]

i would temporary remove this --update-all-extensions

then run it

and try to update extensions manaually

does it crash on startup or during extension update?

if you have it in command line, we will not know what of those things crash

[fading herald]

alright, i'll try without it ! not sure if it's related to my crash but at least it'll solve the controlnet update issue i'm having with the new hand processor

and the crashes seem to be random

[hearty igloo]

there is 2 webui.user. Should i delete 1 tab ?

[fading herald]

just before my question it crashed while i was writing a prompt for instance

[ornate elk]

Close both and reopen one

[karmic crown]

i supect that one of extensions crash it

[ornate elk]

What's your webui version?

[fading herald]

1.7.0 since I've got the git pull, i'm always up to date with the master branch

[ornate elk]

Okay

[karmic crown]

can you do git status, before try to run it?

[fading herald]

in the webui or in a separate terminal ?

[karmic crown]

in terminal (cmd) inside webui folder

❯ git status
On branch master
Your branch is up to date with 'origin/master'.

Changes not staged for commit:
  (use "git add <file>..." to update what will be committed)
  (use "git restore <file>..." to discard changes in working directory)
    modified:   modules/sd_disable_initialization.py
    modified:   webui-user.sh

Untracked files:
  (use "git add <file>..." to include in what will be committed)
    .DS_Store

no changes added to commit (use "git add" and/or "git commit -a")

i want to see what files are modifed or utracked

[karmic crown]

ok, looks fine, nothing important is not messed up

[fading herald]

haven't touched anything except the webui-user, and running the same extensions for months

oh! just remembered, it started crashing after a reinstallation of windows and a full format

[karmic crown]

from the same folder, try this

webui.bat --update-all-extensions

does it crash?

[fading herald]

so maybe i'm just missing a dependancy of some kind?

[fading herald]

still the error about controlnet but no crash

[karmic crown]

ok, go to extensions and check for updates

there should be none

[fading herald]

already did, everything is up to date

[karmic crown]

good

close this

and open normal webui-user.bat

and see if it still crash

now we know that you have updated webui and extensions

[ornate elk]

@hearty igloo did it worked?

[hearty igloo]

installing

[fading herald]

given the crashes seem to be random, it might not crash now, but in 3 hours ou could ^^' I'll see if it does or not and i'll come back here then ! thanks for taking the time 🙂

[karmic crown]

those are the worse, i call them developer's nightmare

[ornate elk]

Can you show the controlnet error?

Maybe you need to reinstall it

[hearty igloo]

its worked bro thanks for help!

[ornate elk]

Perfect! No problem 🙂

[graceful girder]

heyyy yeah its me again lmao so i have a question when my stable diffusion start i can use correctly but when a use openoutpainting i can use theme correctly dans in my terminal i see this (screen) it normal ?

[ornate elk]

Your running an old torch version

[karmic crown]

how do you have pytorch 1.13 again?

[graceful girder]

oh ? shit lmao

[graceful girder]

no i don't think

[graceful girder]

how i can upgrade my torch ?

[vocal fox]

I need some help installing SD 1.5:
when i run webui-user, cmd just opens and closes and nothing happens

how should i troubleshoot?

[karmic crown]

what is your OS and GPU?

[vocal fox]

windows and rtx 3060ti

windows 10

[ornate elk]

What's your webui version?

[graceful girder]

where i can see thats ?

[ornate elk]

Second line in cmd

[karmic crown]

#🤝｜tech-support message <- use this guide

[graceful girder]

okay

is thats ?

[karmic crown]

yes

[graceful girder]

for upgrade my torch i need an other agrs in my commandline ?

[karmic crown]

please post your whole webui-user.bat

[graceful girder]

okay

[ornate elk]

Delete the venv folder

[karmic crown]

ok, thats is not a problem, i was afraid you have something there

[ornate elk]

That should be enough

[karmic crown]

yes

[graceful girder]

okay i delete my venv folder

[ornate elk]

Then relaunch the webui-user.bat

[karmic crown]

when you run webui-user again it will create new venv folder

[graceful girder]

okay okay

[karmic crown]

with new python modules

(proper ones)

[graceful girder]

ohh okay nice

[vocal fox]

I did it with the following info:
@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --xformers --no-half-vae --medvram-sdxl

call webui.bat

And it still doesnt work

[karmic crown]

when you run it, does cmd windows shows? what you have in there? (do not close it or you will close the whole app)

[vocal fox]

cmd blinks open for a second and disappears

[karmic crown]

can you right click on folder here and open cmd here

and than run webui-user.bat, so it will stay open

than show us what you got

[vocal fox]

uh alreight but you know the command for it?

[graceful girder]

i have an other question at the launch of stable diffusions i see no module xformers so what is the utility of xformers ?

[ornate elk]

Xformers is a performance improvement, currently only available for nvidia cards

[karmic crown]

run ti

and show us why it stoped

[ornate elk]

Any third party antivirus?

[vocal fox]

i did, it doesnt do anything

[karmic crown]

thatis strange

[graceful girder]

ohhh ok ok so i can't use it lmao x) thanks

[ornate elk]

Yes, but my guide has the best performance commands for amd included

But xformers would be nice to have ^^

[vocal fox]

nah no antivirus, and i tried turning off my firewall

[graceful girder]

yup it works good for me ^^

[vocal fox]

alright, so what should i do?

[ornate elk]

Restart the PC

Then double click the webui-user.bat

[gusty trench]

Hey everyone - running into issues with SVD through the API. Just getting {"name":"bad_request","errors":["image: invalid format - must be image/jpeg or image/png"]} even though the image is just fine (works via Replicate, for example). Any ideas?

[ornate elk]

still closing after double click?

[vocal fox]

yeah doesnt work still

after reset doesnt work

so what now

[ornate elk]

Did you run at is administrator ?

[vocal fox]

ya

[terse pebble]

Hi guys, new to sd discord. I have an issue. When I generate with deforum I don't get the same result as tests in img2img. The results are pretty much different. I am 99% I copied all the settings from IMG2IMG to Deforum, including control net, cfg, steps. What I am missing ? first img2img second deforum

[karmic crown]

denoise?

[terse pebble]

denoise in img2img or deforum? where I can find denoise in deforum, maybe I skipped that. In img2img it's 1 because I use open pose, tile, temporalnet with controlnet for my consistency results.

[karmic crown]

i am currently playing with img2img and i am going crazy trying to find a proper denoise to get what i want, so that is the first thing that popped to my mind

[terse pebble]

I always use my denoise to 1 in img2img due to controlnet, but maybe that's the problem. I have to search how to set the denoise strenght on video deforum

[ornate elk]

Then run without

[ornate elk]

Denois 1 will change an image to 100℅

[terse pebble]

how can I set the denoise strenght on deforum video to video ?

[vocal fox]

neither works bro

[karmic crown]

i am confused why you do not get ANYTHING

so, you removed your old installation and followed guide that i showed you?

[vocal fox]

oh it was the exact same method i did before, except renaming the .bat file

but i can try removing my old installation and redoing it i guess

how do i "uninstall" it? a git command?

[karmic crown]

please follow this guide #🤝｜tech-support message

[karmic crown]

ok

first

before you do anything

open cmd window

and try those commands
python --version
pip --version
git --version

did all of those work?

[vocal fox]

older version of python i guess is installed

[karmic crown]

ok, git is fine

leave it

[vocal fox]

is there a reason why these guides dont just have you install a venv?

i need python 3.10 right?

[karmic crown]

yes

venv will be created by webui when you run it

come to the dark side (mac) we have script (i made it) 🙂

please install python 3.10 as described, leave git as it is

[vocal fox]

i'm sure its on this computer

[karmic crown]

remove pyenv (unless you are a python developer)

[vocal fox]

i like to use anaconda sometimes for coding...

must i break all my other stuff to make this work?

[karmic crown]

no

if you use pyenv for something else, leave it

[vocal fox]

ok

[karmic crown]

but, you might need to specify python path in webui-user.bat later

thats why i told you to remove other python versions if you do not need them, since it would be easier if you have only one

[vocal fox]

yeah i'm gonna try to keep my python installations

[karmic crown]

i have 10 of them

🙂

[vocal fox]

alright so what now

do i uninstall my older version of sd?

how do i do that? just delete the folder?

[karmic crown]

just delete the folder

it is enough

uness it made some strange changes in your global version of python, but since you didnt have 3.10 its fine

[vocal fox]

doesnt work

[vocal fox]

same issue

[karmic crown]

ok

open cmd

and type

python3.10 --version

does it work?

your default python is not 3.10, and that is the problem

we will fix it

first, let me know if this works

[karmic crown]

have you checkt add to path when you installed it?

[vocal fox]

hmm maybe i forgot

[karmic crown]

this

run installation again, and be sure to check this

then open a new cmd and try again

[vocal fox]

its working! thanks

yeah sorry for the msitake there

[karmic crown]

ok

please show me
python3.10 --version

is it working?

and
where python3.10
i hope this is the command for windows, since i am using windows very rarely in a last few years

[willow parrot]

after updating pytorch and installing xformers, i cant generate images anymore

i get 2 errors while booting up the a1111 webui and when i try to make an image i get a very long error

[ornate elk]

can you show the erros, also whats your gpu?

[willow parrot]

I'm already uninstalling everything hoping it works

using a 3060

[ornate elk]

Follow my install guide from the Pinned Messages in this Channel

The one for Nvidia

[willow parrot]

sure

[steady thistle]

@ornate elk I seem to be having less GPU memory the more pics I produce. I get errors about not enough memory even if I had been making larger pixel images earlier.

I started using --lowvram but I dunno if I'm missing something

And having 12GB graphics memory, this is strange.

[ornate elk]

what model are you using ? also what resolution?

[untold marlin]

So question is there some special settings or something to use or a guide somebody can point me to on trying to get it to generate tile sets for game assets?

[steady thistle]

Size: 712x712, Model hash: 8557e300af, Model: afroditexlNSFW_20, Version: 1.7.0... This is now working after putting --lowvram

but gave an error with --medvram

[ornate elk]

is it an sdxl model (6gb) ?

[steady thistle]

It's an XL model, yes. And it is 6.77gb.

[ornate elk]

Yea these heavy models are gonna get you out of vram in the directml version

--lowvram can help

But I would recommend to use 1.5 based models, they are only 2gb and can be easy upscaled

[steady thistle]

Ah, okay. Does --no-half --disable-nan-check in the args hurt anything? I need that for image to image face swapping.

[steady thistle]

I haven't figured out how to upscale yet. 😛

[ornate elk]

Remove --disable-nan-check, that doesn't do anything other than making problems.
--no-half is needed for AMD cards for using inpainting or face swap. But will increase the vram usage by a lot

[steady thistle]

Thanks!

[ornate elk]

If you need help feel free to ask

[steady thistle]

Well, I do hires fix but it just breaks over 1.3 or so. Is there a way I can upscale without killing my vram?

[steady thistle]

I don't know much about this but if you need any US legal help, feel free to ask me. That's something I do know. lol

[untold marlin]

Can stable diffusion even do tile maps cuz no matter what I put in the prompt I seem to just be coming out with like construction tiles that are very pretty

[ornate elk]

Yes with the tiled vae extension

[ornate elk]

You mean tiles for a Texturepack ?

There is an setting for that

[untold marlin]

I mean something more like that where you can mix and match pieces to create worlds and stuff like that

[steady thistle]

I'll experiment with that, I haven't seen it before and don't really know how to use it.

[ornate elk]

Its easy, you only need to enable tiled vae.
Then set the Encoder to 1024 and the Decoder to 128.
Then you are able to upscale by 2 from 512x768 or from 540x960 to get FullHD. Make sure to set Hires steps to 10. Denois to 0.5.

Also this works only for 1.5 models

[steady thistle]

Ah okay, very useful stuff. I really like this Aphroditexl checkpoint but I'll play with it. What's a good 1.5 checkpoint for realism?

And just to be clear. I can create in those resolutions, get an image I like, keep the same prompt and seed then turn on the tiled vae and hires with those parameters and it'll work out?

[ornate elk]

no you have to enable tiled vae, whenever your using hires fix

or whenever your resolution goes over 1000p

thats how it will work

sdxl can work with hires fix but only when using 768 and upscale by 1.5

but i havent tested that on 12gb amd

if you want to upscale larger to like 4k you need to use the sd upscale script in img2img, let me know when you want to try, ill guide you if needed

[opaque viper]

Help please. For no reason at all, my generations are suddenly getting deep-fried. It's worse with some samplers (dmp++ sde) and worse at higher cfgs (even 6 or 7). I didn't change anything

[ornate elk]

your image doesnt get upscaled, but your using an upscaler

[opaque viper]

What do you mean? The upscaler is later in the workflow

[ornate elk]

okay i just checked the meta data

[opaque viper]

Weird, I'll try disconnecting everthing from the upscale stuff

[steady thistle]

Thank you! I'll do that! I'll push and see what I can do with ReActor, sdxl, --medvram and 512x768. First, I forgot to hit tiled vae and it died at the ReActor stage. Then I turned on tiled vae properly and that died. Then I set it to --lowvram and ran it through the same way and it worked. Can confirm, with 12gb, you can use a 6gb checkpoint with tiled vae.

But it is not fast.

[ornate elk]

yea --lowvram makes it slower and also sdxl in general is slower

but great that it worked

[steady thistle]

What do you think of the quality tho? Looks grainy around the edges? Is that my prompt and poor negatives?

[ornate elk]

thats the latent upscaler

use latent bicubic

or Resrgan4x+

[steady thistle]

Aah, okay.

[steady thistle]

Okay, how do I change that...? Can't find it...

[ornate elk]

upscaler in hires fix

[opaque viper]

So the culprit seems to be the negative prompt. I took it out and it became fine

[steady thistle]

Oh, duh, I didn't have the menu dropped down in hires fix. I'll try the same image with Resrgan4x+

[ornate elk]

thats strange

[steady thistle]

I probably messed something up but I think it's moving right this time.

Yeah, much better!

[kindred tinsel]

Hello! Is there anyone around who can guide me through again sorry on the installation for stable diffusion for AMD cards, I forgot how to do it and I want to reinstall that and comfyUI. I would much appreciate it ^^

[ornate elk]

hey, checkout the Pinned Message in this Channel, there is my guide for AMD

[kindred tinsel]

POh I see also hi! Also, do you remember you told me you could help[ me out set the drives such that, the m2 had the webuis and the hdd had the models without affecting perforamnce that much/

Well I got a new 1tb ssd an have the models on a 4tb hdd and was wondering if you could help me out to set up that simlink or smth

[ornate elk]

sure

[kindred tinsel]

okay let me get python, git and the other thing done and Ill ping you for that okay? :)

[ornate elk]

its 2am here, i might be off in like 20 mins

[kindred tinsel]

oooof, well we can do it later if you'd like, mind telling me what your timezone is?

Uh Im not sure if this did everything that the command needed? @ornate elk

[ornate elk]

its done

[kindred tinsel]

ok ok im installing but it said it did not recognize --no-half as an internal command

, im going to see if it's because of some kind of space I put in or smth

[ornate elk]

yea is mostly because of space

[kindred tinsel]

oh if you are up would you mind helping me out with the symlink, im confused :c

everything works fine in the app and the only thing I am missing is the models :)

[graceful girder]

hello yes it's me again but this time I come by that I don't understand the reason for the stable diffusion crash when I try to do inpainting and outpainting with control net someone could enlighten me ?

[ornate elk]

do you want to use cmd to make symlinks or do you want a handy tool?
Commandline:
https://www.educative.io/answers/mklink-windows-10
Tool:
https://schinagl.priv.at/nt/hardlinkshellext/HardLinkShellExt_X64.exe

[steady thistle]

@ornate elk Sorry for all the questions. I'm trying to use ReActor to replace faces from old movie stills. What's the best way of doing so?

[kindred tinsel]

I wanted to know if I should wrap the hardlinks like D"/smth/smth/smth/models/Stable-Diffusion with ""

I think I can manage with the link you send I am confused about the "" :)

[ornate elk]

oh with that i cant help, reactor replaces the face while generating, idk if it will work in img2img

[steady thistle]

No worries, I think I have it but figured I wasn't being very efficient about it.

It does work, tho. Just run the photo without prompts, make sure the size is unchanged, have the face replacer, and SD simply processes the image with no changes then ReActor kicks in.

I think. lol

I got it to work before my big crash and don't recall for certain

[kindred tinsel]

Well I did the connection using the link shell extension @ornate elk , it appeared to have done something as it asked me for admin permission, but the models are not connected :c I can only see the pruned models

basically sd1.5 downaloded by webui ;c

[ornate elk]

you need to make a symlink with the tool

[kindred tinsel]

I did, but I think I did it the other way around, I pressed the make symlink link like the button on the new installation which is empty, which effectively did nothing s

[ornate elk]

you need to right click the models folder on D first

select it as source

then right click in the empty models folder from auto1111 and select destination

[kindred tinsel]

I got it !!!!!! Yeah I notice what you mean it was my error yeagh, not it detected all of the models yeye

It was the other way around first full then empty, so I will do that with every other folder in there too ^^ thank you !

[ornate elk]

you can do that with the whole models folder and lora, the others are not needed to move on D

[kindred tinsel]

Oh really? but I could do it with embeddings and the other too right?, Since I am going to run comfy and it allows me to define folders I figured if I connected usign symlink to webui then linking those models to comfy would transfer the link

So I wanted to do everything at once so there were no issues, also, im not sure what that mean

[ornate elk]

yes you can do that too

[kindred tinsel]

oh nice! but im scared about that error, it was working a second ago

[ornate elk]

For Comfyui:
After downloading and extracting it you need to find the the file called: extra_model_paths.yaml.example, Rename this file to extra_model_paths.yaml and edit it with your favorite text editor. Then at the first line set your path to Auto1111 webui folder. then save. Comfyui now has access to all models from auto1111

[kindred tinsel]

yeye ^^ but that ADL_2 error, what is that? :===

[ornate elk]

never saw it before

[kindred tinsel]

FIXED IT

okay so

The issue was that I changed some method for a111 to get performance counters in vram, using the amd optimization only, the one that says (amd only) but it kills the install and it is a known issue haha, sorry

[ornate elk]

oh lol xD

[kindred tinsel]

Its just the typical AMD shit

basically stuff doesnt work with our cards deal with it

Does anyone know why this failed for segment anything and inpaint anything? It says ti doesn't have permissions but I am not aware I needed to open the webui with admin permissions for installing extensions or the like

[kindred tinsel]

I tried reinstalling the webui but it doesn't work at all :c Im not sure what I am supposed to do to install these extensions without getting those errors

Happend this time with controlnet ;c Im guessing that this new requirement might also not work at all so idkl...... please help :c

[sonic ferry]

I installed Pythongosssss Custom Scripts and have been having trouble, first I swapped out the Model Loader and Lora Loader for his, got an odd error with the Model Loader boxed in red. Eventually I figure out that it doesn't like the original prompt boxes I had, accidentally found this out when I hit something that automatically gives you a positive and negative prompt from his Model Loader. Looks the same but get rid of mine and connect everything back up, got a picture to generate. Start playing around with adding a chain of LoRA...so far so good, but then after a restart of ComfyUI it's giving me an error again. Doesn't like the LoRA anymore, remove Loaders and add them back...it's working again. But now my whole PC is crawling, a 30 min CPU gen is now saying hours, can't even switch tabs anymore. I eventually close the ComfyUI cmd window and open it back up, more issues, now the Model Loader reset to nothing, put it back on EpicPhotogasm, now it's giving me an error again, do the Prompt boxes thing again, nope nothing. Anyone use this thing? I like that I can see descriptions of what I'm using, add notes instead of using Notepad++ like I'm doing now, but it was falling apart very shortly after I thought it was working great.
Is there something I'm supposed to do to get it working right?

[woven smelt]

Hello I am trying to train a lora with koyha ss for first time. I have watched a few youtube tutorials but cannot get my seocnds per iteration to reasonable level. with 32 network dimension its 14 seconds, 128 is 30 seconds etc. I have 2080 ti, 12gb memory pls 16 shared memory. I can post the config if it helps but not sure what i am doing wrong

Since youtube videos I watch have 2 it/s instead of 30 s/it

Even with 32 network and 512,512 its 12 s/it. Training sdxl btw

[woven smelt]

added --network_train_unet_only and it dropped to under 2s/it with 1024,1024 and 32 network

[woven smelt]

Although still not sure if thats as fast as it could be, anyone else training loras and have their card and speeds?

[tulip mesa]

Trying here again. It takes forever (2 min) to use wd 14 tagger or rembg in stable diffusion (Because i think I've updated to tensor RT). Does anyone know how to fix this? Or remove tensor RT?

I found this link, but have no idea if it works or if it's even safe. https://github.com/Haoming02/WD14-Tagger-TensorRT

[sweet basin]

everything was fine but one day I tried it and keep getting that torch can't use gpu.

[ornate elk]

To remove tensorRT, delete its extension from the extensions folder.
Then delete the venv folder

[ornate elk]

What's your GPU?

[tulip mesa]

I'm using comfyui, can't really find the extension folder

[ornate elk]

Oh okay

Idk where comfyui stores these

[tulip mesa]

k. I'll give up i guess and restore my computer fully instead 😅

[ornate elk]

You could just reinstall comfyui

Why the whole PC xD

[tulip mesa]

isn't tensorrt installed in the graphic driver? cuda or some shit like that?

[ornate elk]

Nope, I guess you only installed the extension and didn't manually changed stuff in the Nvidia control panel

Then everything is fine

[tulip mesa]

no, i installed it on cuda and updated environment variables and stuff like that

[tulip mesa]

I followed this guide

[ornate elk]

Okay but that is reversible too

[tulip mesa]

downloaded a new copy of comfyui, working now. Thanks

[ornate elk]

Nice np

[sweet basin]

6800 xt 16gb

[late tree]

Anyone need a reminder not to click suspicious links?

[ornate elk]

Then you need to add --use-directml --medvram --opt-sub-quad-attention --opt-split-attention-v1 --upcast-sampling --no-half-vae
To the webui-user.bat at the line COMMANDLINE_ARGS=

[sweet basin]

it worked fine for a year

et COMMANDLINE_ARGS= --medvram --no-half --precision full --no-half-vae --opt-sub-quad-attention --opt-split-attention-v1 --disable-nan-check --upcast-sampling

my current

[ornate elk]

Remove --precision full --disable-nan-check
And add --use-directml

Thats needed for amd since webui version 1.7.0

[rotund lava]

try to add this argument --disable-model-loading-ram-optimization

then restart webui again

[slate shore]

add or replace current

--precision full --no-half --disable-nan-check --medvram --xformers

current^

[ornate elk]

What's in your webui-user.bat

[rotund lava]

add to the end

[ornate elk]

Try to delete the venv folder

[sweet basin]

I deleted few days ago but can do it again

[ornate elk]

Remove --disable-nan-chexk

[slate shore]

should it be opening 2 cmd terminals on clicking the webui

[rotund lava]

u have to close the cmd to restart webui

[slate shore]

its not errored yet.

[rotund lava]

now u got it working,when its done tell me how long it took

[slate shore]

its going faster but not really usbale i dont think and i think its going to be a way messed up image

[rotund lava]

that one should take like 1/2mins

[slate shore]

create a hyperrealistic image of a lime green lamborghini countach in a cyber punk background
Steps: 20, Sampler: DPM++ 2M Karras, CFG scale: 7, Seed: 1932712431, Size: 512x512, Model hash: cc6cb27103, Model: v1-5-pruned-emaonly, Version: v1.7.0

Time taken: 3 min. 48.1 sec.

that is SOOO bad lol

[ornate elk]

Remove --precision full and try again

[rotund lava]

yea it looks bad cause u are using the default 1.5 model

[slate shore]

i think ill just wait till the 4070 gets here ... we like doors to go up not float up haha

removing --precision full

[rotund lava]

to get better results download any of the popular models from civitai

[slate shore]

oh you have to start with a model

ive only made it throught he installation videos, not dove into any use case videos

[rotund lava]

you need to download a custom model and put it in models folder then select it from dropdown list to get better results otherwise it will use the default one

[slate shore]

so you find a model in there that kinda has the style you are going for and then modify that

is it advisable to go more toward the exact object type or image style... EG if i want a specific car in a specific scene is it better to find a scene model or a car model

create a hyperrealistic image of a lime green lamborghini countach in a cyber punk background
Steps: 20, Sampler: DPM++ 2M Karras, CFG scale: 7, Seed: 2541744399, Size: 512x512, Model hash: cc6cb27103, Model: v1-5-pruned-emaonly, Version: v1.7.0

Time taken: 3 min. 40.5 sec.

--no floating doors this time but it looks more like a lotus than a lambo

[ornate elk]

Download Dreamshaper v8 and use Euler A as sampler

It will be better output and should be faster

[rotund lava]

the popular ones can output almost anything but there are some models that are made specifically for certain style

[slate shore]

i will do some testing, I genuinely appreciate the help guys. Need to cook some breakfast for little one... ill dive in a bit later and test dreamshaper/euler a and snag some models

[sweet basin]

thx worked

[lean ridge]

hey, im trying to use the "resize by" tab at x2 from 768x1344 to 1536x2688 but in the last 99% im getting cuda out of memory. however i have achieved it before. there's a way i can get the render done?

using XL

I have a rtx 3060 12gb

set COMMANDLINE_ARGS= --xformers

[ornate elk]

Hey for upscaling you should either use hires fix in txt2img or SD upscale script in img2img

[lean ridge]

but it's not possible on that tab?

[ornate elk]

Its not possible with resize when using to high resolution

Because your trying to generate at 1536x2688

[lean ridge]

nvm, just done it at 1.5, ty!

[fading herald]

what's the correct syntax to use --ckpt-dir and the other arguments for custom folders? I can't make it work for some reason

I tried ```
--ckpt-dir "E:\Checkpoints"
--ckpt-dir E:\Checkpoints
--ckpt-dir="E:\Checkpoints"
--ckpt-dir=E:\Checkpoints\

[rotund lava]

remove the last \

--ckpt-dir "E:\Checkpoints"

[fading herald]

oh come on, really, just that ? x) Seems to be working, just have two unrecognized arguments, which is weird since I'm pretty sure I've seen them before, --esrgan-dir and --controlnet-dir

it's --esrgan-model-path for the first one, but can't find the argument for controlnet in the list, maybe it doesn't exist then...

[rotund lava]

i think controlnet is no longer used cause u can go to settings>controlnet and set a different path there

[fading herald]

oh didn't know that! thanks for the tips 🙂

[kindred tinsel]

Does anyone know why this failed for segment anything and inpaint anything? It says ti doesn't have permissions but I am not aware I needed to open the webui with admin permissions for installing extensions or the like

Hello I was hoping now there might be someone out here that might be able to help me out ^^.

Basically I have q python installation for all users, so is installed In c/ program files.

I have my webui in c/Stable Diffusion Editors/ sd-webui-directml

Here I open up the vent correctly and prerequisites are installed with no issue, but when it comes to installing extensions they are getting blocked by those permission.

Does it have to do with the way I installed python and where I did? The folder for stable diffusion wouldn’t be an issue right?

[burnt cliff]

Guys, would you know how to solve that ?

[kindred tinsel]

@ornate elk do you know anything about fixing this bug perhaps?

[woven smelt]

If I create a lora and I decide it needs further training can I start it off from the epoch it finished at or do I need to fo the whole thing again?

[hearty igloo]

do i need to click launch to launch stable diffusion ? i cant see any run options

[tame yoke]

is there a branch or something for gpus that dont pass cuda torch test

cuz old

[karmic crown]

wait, do you mean on amd on windows, or any system?

[hearty igloo]

do i need to click launch to launch stable diffusion ? i cant see any run options like run.bat

[naive remnant]

I'm facing this error while installing for the first time.
Can anyone help fixing it ?

[karmic crown]

that was a question for CS1o, about Tiled VAE, lets see whats bothering you now 🙂

[kindred tinsel]

Does anyone know why this failed for segment anything and inpaint anything? It says ti doesn't have permissions but I am not aware I needed to open the webui with admin permissions for installing extensions or the like

[karmic crown]

if you are on windows, you run automatic1111, with webui-user.bat, and if you are on windows on mac with webui.sh

[kindred tinsel]

I posted the details on a higher up post if anyone wants to look at ir ^^

[karmic crown]

if you run anything as admin, you might corupt your folder privileges

[kindred tinsel]

I can’t run as admin

[hearty igloo]

I just have those

[kindred tinsel]

It doesn’t open the webui file

[karmic crown]

my point was - do not even try to run as admin

[kindred tinsel]

Yeah but I don’t run anything as admin, even if I wanted to I cant. I’m not sure if the issue stems from the folder being in a standalone directory and not in my users folder in my C drive

[simple tartan]

anyone else here running both SD and text-gen? even with different ports, I can't seem to have both running at the same time

[wet surge]

Can sb help me? I am trying to install and set up the ai, but i don't have pytorch_lightning and I cann't install it

[karmic crown]

some of your folders obviously have some other privileges than normal, and you cant write on those folders as an user. things like those happens when you run something as admin, and that admin command change privileges of some of your folders

[kindred tinsel]

Okay? So what should I do then, install python as only for one user and keep the same directory ?

[karmic crown]

i do not have access to windows currently, so i cant help you much, but in cases like those you need to go to your top folder and change the permissions for that folder and all folders inside

[kindred tinsel]

In my precious install I had my python installed with normal for one user and was able to run everything from another drive not even in c. So imma try this out see if that helps as I assume that most requirement installs are done for the -m

[karmic crown]

if you were on linux or mac i would give you a smiple terminal command for that

it is not a python permission

it is something in your automatic1111 installation

[ornate elk]

On AMD on windows or when you get grey squares

[karmic crown]

@kindred tinsel

it is something in your venv folder

[ornate elk]

Wrong python versjon

[karmic crown]

so, just remove venv, and let webui create a new one

[ornate elk]

What's your GPU?

[kindred tinsel]

Yeah but it says permission denied which might be because of the way I installed python

[kindred tinsel]

Did that twice yesterday and none of it worked

[tame yoke]

gt 430 under a cover of 730

[ornate elk]

Won't work with SD, you can only use CPU mode

[vocal burrow]

there seems to be error even before that, treat them one by one in order

[ornate elk]

Run the webui-user.bat
The second from below

[kindred tinsel]

Basically it happens for most extensions installed directly through the webui

No prerequisite for the normal webui failed just those for extensions

[vocal burrow]

then remove them, and try to install them one by one. fix the error as they come

[kindred tinsel]

I tried that, and it’s always about that cv2.pyd file. I even tried installing controlnet from GitHub and nothing .

Thats why I push this as a permission error rather than dependencies or folders. I might just try reinstalling python for just one user and try again

[hearty igloo]

thanks man

[vocal burrow]

you might have but I didn't see any of that. All I'm seeing is :

• incomplete logs
• multiple errors on top of each others meaning it's gonna be a clusterf*** to fix all at once

[tame yoke]

is there a way to make a cpu pass the "hello asuka" test then?

[kindred tinsel]

Here are some other screenshots of what happened last night

[kindred tinsel]

What logs do you want to see the app opened and I threw those errors

[vocal burrow]

it shouldn't matter

[tame yoke]

it does tho

[vocal burrow]

ok then what's the problem ?

[hearty igloo]

downloaded juggernaut model and got this error what is this ?
module._apply(fn)
[Previous line repeated 5 more times]
File "C:\Ai\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 820, in _apply
param_applied = fn(param)
File "C:\Ai\stable-diffusion-webui\venv\lib\site-packages\torch\nn\modules\module.py", line 1143, in convert
return t.to(device, dtype if t.is_floating_point() or t.is_complex() else None, non_blocking)
NotImplementedError: Cannot copy out of meta tensor; no data!

[tame yoke]

it doesnt output the same image as it would on a gpu

[vocal burrow]

like I said uninstall every problematic extensions, install one, relaunch, take the whole console log (put in in a .txt) and drop it there. We'll see from here.

[kindred tinsel]

Would you mind if I dm you with that info a bit later, I’m a little busy out from home and this chat is chaotic

[vocal burrow]

That's not very descriptive, what's the image ? what are your settings ? what does the log says ?

[vocal burrow]

you can ask here later on. I don't usually answer to DM (I get too much nowadays).

[karmic crown]

@ornate elk @vocal burrow do you have any screenshot how to change privileges to top folder and all folders inside on windows, he has some folders with wrong privileges which is the source of all 99% of his errors

i do not have access to windows right now

[lone hemlock]

Hello team, I followed the steps in this tutorial https://hub.tcno.co/ai/stable-diffusion/automatic1111-mac/ : and I have this error "*** Arguments: ('task(2utc38y9r54wo23)', 'dog\n', '', [], 20, 'DPM++ 2M Karras', 1, 1, 7, 512, 512, False, 0.7, 2, 'Latent', 0, 0, 0, 'Use same checkpoint', 'Use same sampler', '', '', [], <gradio.routes.Request object at 0x17fbd9330>, 0, False, '', 0.8, -1, False, -1, 0, 0, 0, False, False, 'positive', 'comma', 0, False, False, 'start', '', 1, '', [], 0, '', [], 0, '', [], True, False, False, False, 0, False) {}
Traceback (most recent call last):
File "/Users/censuredusername/Documents/stable-diffusion--webui/modules/devices.py", line 150, in test_for_nans
raise NansException(message)
modules.devices.NansException: A tensor with all NaNs was produced in Unet. This could be either because there's not enough precision to represent the picture, or because your video card does not support half type. Try setting the "Upcast cross attention layer to float32" option in Settings > Stable Diffusion or using the --no-half commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check." Any ideas? I've also tried adding arguments to webui-user.sh like '--skip-torch-cuda-test --disable-nan-check " but now it's python that's crashing. I have Macbook M1.

[tame yoke]

give me a moment

[vocal burrow]

https://www.addictivetips.com/windows-tips/change-file-permissions-windows-10/ (same steps for a folder). I'm not sure that will fix it tho, permissions error can happen for a bunch of reasons.

[kindred tinsel]

I will ping you here then in about an hour once I get home, thanks ^^

[vocal burrow]

would need the full log, put it in a .txt and drop the file there

[karmic crown]

@lone hemlock so, your installation is in Documents/stable-diffusion--webui?

[karmic crown]

i am pretty sure he run something as admin, and f...d up the permissions with that

[lone hemlock]

yes 😉

[karmic crown]

are you sure it is --? and not Documents/stable-diffusion-webui

[vocal burrow]

probably better to start from scratch then. That way perms will be set correctly for sure.

[karmic crown]

i agree, but... 🙂

[lone hemlock]

yes look : "censuredusername@MacBook-Air-de-censuredusername-2 stable-diffusion-webui % pwd
/Users/censuredusername/Documents/stable-diffusion-webui
censuredusername@MacBook-Air-de-Jeremy-2 stable-diffusion-webui %
"

[kindred tinsel]

Bro I reinstalled the webui TWO times,

[karmic crown]

so i was right 🙂 /Users/censuredusername/Documents/stable-diffusion-webui (just only -)

[kindred tinsel]

Didn’t run anything as admin because there is no need for a bat file to do that, why would I raise permissions where there is no mention of that

[karmic crown]

ok, download my script https://github.com/viking1304/a1111-setup/releases/tag/0.0.5

[kindred tinsel]

Literally only installed following the guide from Cd here

[wet surge]

How can i download this torch? It doesn't got downloaded form the requirements.txt

[vocal burrow]

full log please (put it in a .txt and drop the file there)

(I should write a macro to copy paste this .... And as usual nobody follow up on that :D)

[karmic crown]

and run it like this
sh a1111-setup.sh -f all -d Documents/stable-diffusion-webui

[lone hemlock]

the script is running

[kindred tinsel]

Okay since it is such an issue, I am going to

1. Reinstall python only for my user,
2. Reclone the git repo for stable diffusion
3. Try to Install controlnet again to replicate output

I will keep the same directory structure since I am sure that creating a folder in the C drive doesn't mean the end of the world

[karmic crown]

it will just fix a1111 instalation and command line argumments, but it will not touch your models or extensions (if you downloaded any)

[kindred tinsel]

Could this be the issue @karmic crown ?

Or ticking the first box there be the issue?

[vocal burrow]

incompatible python.
You need to install any python from 3.10.6 to 3.10.13 (only up to 3.10.11 is available as installer for windows)

[karmic crown]

i would remove the first checkbox, but that is question for windows guys

[kindred tinsel]

okay calling windows guys

[vocal burrow]

leave everything as default except for the "add to PATH" and eventually the install folder

[kindred tinsel]

So I shouldn'[t install it for all users? Sorry but what does it do that messes up installations?

or rather, what does that mean, does it give it admin permissions or smth?

[vocal burrow]

install for all users will install it in some folder accessible to everyone thus it requires some special permissions, but it shouldn't matter for SD as it's only gonna read files from that folder and not modify anything in it

[lone hemlock]

You are a champion !!! it works thank you very much !!! you should talk about it on reddit because I had found nothing as info for mac M1

[wet surge]

Like this?

[vocal burrow]

is that the microsoft store ?
if so toss that into the garbage bin. I wouldn't trust windows to not update python behind your back without telling you.

[kindred tinsel]

Okay, well I did as you said, leave everything as default since the python gods that wrote the installer know best, Imma go clone the repo now

[wet surge]

no

[karmic crown]

i am glad i could help

[vocal burrow]

Just take it straight from python's website https://www.python.org/downloads/release/python-31011/

[kindred tinsel]

What is the command to update the pip, i might have a low ass verison 23.0.1 when some of the screnshots said there was 23.3.2 or smth out

[hearty igloo]

which one is most close model to mid journey?

[karmic crown]

forget that, its fine

[kindred tinsel]

Okay then

[vocal burrow]

it should give it to you (in green) whenever you launch sd

[steady thistle]

Are the 6gb models XL models? And the 2gb 1.5 models? What would the 4gb be?

[kindred tinsel]

Alright this is the new install for python

[tame yoke]

comes out like this (same exact settings)
it does not pass cuz its different from original image

[karmic crown]

looks good

[vocal burrow]

it's not that easy, 1.5 models can be 2, 4, 6, 8gb or even more. However you're probably not gonna get any SDXL model under 8gb.

[kindred tinsel]

not as straightforward, there are 6gb sd1.5 models (unpruned), 1.99 gb sd1.5 models, 1.99 sd2.1 models, etc

[steady thistle]

Hmm... Will the SDXL models all have XL in the name somewhere, generally speaking?

[kindred tinsel]

Look this is my c drive, this souldnt cause any problems right?

[karmic crown]

my advice do not use spaces

[vocal burrow]

generally yes, people always indicate what version they use as a base be it in the name/description/etc

[kindred tinsel]

Letme try that

[karmic crown]

most things can handled that well, but better be safe than sorry

[vocal burrow]

I wouldn't use any space just to be safe

[kindred tinsel]

I changed it to StableDiffusionIDES so its all closed together, now according to the guide pasted in here, I should do this command git clone https://github.com/lshqqytiger/stable-diffusion-webui-directml && cd stable-diffusion-webui-directml && git submodule init && git submodule update

I am not sure if this is accurate though since on my last installs I litertally only cloned the model and that it,

[proper oracle]

Morning y'all. I'm still in the early phases of my learning journey for training dreambooth lora models. Anyone have some good resources they can point me to for overfitting/over training? Specifically, I am trying to get my Lora to work well at a strength of 1, currently I need to drop to between .6-.8 to get a satisfactory result which ends up losing some of the detail of the subject.

[karmic crown]

yes, but be sure to remove or rename old stable-diffusion-webui-directml folder first

[proper oracle]

A few things I'm considering, 1. quality of my images are poor. 2, my learning rate is too fast. 3, the class i'm training against is too broad (thus resulting in all renderings of subject across that class).

[kindred tinsel]

Folder do be empty so, command shouldn't cause problems right?

@karmic crown :)

[karmic crown]

ok open cmd from there and paste those commands

[kindred tinsel]

Each of them on their own or the entire line at the time?>

[karmic crown]

everything at once

[kindred tinsel]

Itdid this, but I am not sure if it executed the rest of the commands

[karmic crown]

looks fine

[kindred tinsel]

alright so now Imma go ahead and run the webui file with the command line args provided here and update on my results :)

[karmic crown]

show us your webui-user.bat after edit before you run it, just to be sure

[kindred tinsel]

Would this be something to use in this case since the permissions erros arose?

[karmic crown]

just add what is in the guide

[kindred tinsel]

Light mode bomb

[karmic crown]

remove this space, you do not need it

[kindred tinsel]

damn okay

Looking like this now

[karmic crown]

remove --no-half-vae

[kindred tinsel]

Cant do

that is a requirement for amd cards

as is --no-half, or else it dies

[karmic crown]

you do not need both, just one of them

but in order to simplified the guide, CS1o, didnt go into details

[kindred tinsel]

BRO

Ive got a RX 5500 XT :c

the reason that line was added was because of the constant problems my card series was giving lmao

[karmic crown]

will it work with boht? yes. do you need both? no. 🙂

as i just explained guide would be too complicated if he goes into those details

[vocal burrow]

show your settings, I just confirmed that it's still working

[kindred tinsel]

Idek my man, Im not too knowledgeable in this Ai thing, rn I know it says add both, then I add :)

[vocal burrow]

--no-half <=> don't use half type for the whole SD pipeline
--no-half-vae <=> don't use half type for the VAE part of SD pipeline

[karmic crown]

personally, I would writ this
RX 5xxx cards: instead of --no-half-vae, use --no-half

[vocal burrow]

so you don't need --no-half-vae if you use --no-half already

[kindred tinsel]

but rn I am running the install and no errors seem to be arising, it installed all default pacakges, cloned stable diffusion, cloned stable diffusion xl, cloned k fissuion, cloned codeformer, cloned blip, and not is installing requirements for codeformer and general requirements

[kindred tinsel]

Fine jeeez, Ill change that once this is done

[karmic crown]

as we said, it will work with both of them, but you do not need both 🙂

[kindred tinsel]

It do be taking its time with the requirements hah

[karmic crown]

but CS1o tryied to make guides as simple as possible

[vocal burrow]

depends of your connection and hardware

[kindred tinsel]

Ye ye Imma have to get better wifi lmao, but in any case its downloading the base model cus, bleh it has to do tht every time sadly. Rn its at 55% once its done Imma try installing ONE extension, control net see if the issue arises again

[karmic crown]

i just timed my script, it took 3:04.95 to install everything and download 1.5 model 🤣

[vocal burrow]

yeah it probably takes me less than 2 minutes. But I've got fiber connection and 12 cores cpu.

[karmic crown]

1Gbps? I am on 600/60Mbps

[vocal burrow]

theoretically yes

[kindred tinsel]

My man I got I think 14mbs fastest download speed when my internet says its a 60mbps something speed

but

@vocal burrow

[karmic crown]

i am getting 585/58 Mbps over wifi 🙂

[wet surge]

I have succesfully installed the pytorch_lightning packages, but it still not working

[kindred tinsel]

That is the log for the installation, as you can see nothing went wrong with it, all of the initial dependencies including torch and torchdirectml were installed normally

[vocal burrow]

looks good to me

[karmic crown]

fatal: No names found, cannot describe anything.
wtf?!

[kindred tinsel]

noraml thing tbg

[vocal burrow]

yeah it's normal for directml fork

[kindred tinsel]

always says that when installing

[vocal burrow]

they didn't clean the script from upstream

[kindred tinsel]

This is the extension I am about to try installing, wish me luck

[karmic crown]

in that case, i have no complaints

[vocal burrow]

what are you trying to use ? why are you using python 3.12 ?

[kindred tinsel]

Nope

errors

again

[vocal burrow]

kay now that we have a clean base, show us the full log

[kindred tinsel]

its always that fricking .pyd file wtf is that

[wet surge]

In the cmd It shows that the version is 3.10.11

[kindred tinsel]

wait its trying to isntall trimesh[easy], should I let it finish and then send in?

[karmic crown]

do you have som strange antivirus?

[vocal burrow]

and in your log it says "C:\Users\AMD\AppData\Local\Packages\PythonSoftwareFoundation.Python.3.12_qbz5n2kfra8p0 3.12

also again, what are you trying to use ?

[wet surge]

I guess I had 3.12 before

[vocal burrow]

sure

[kindred tinsel]

I just installed windows, it only runs on normal windows defender for now, not antivirus software is installed

[vocal burrow]

then you're probably using an old venv from back then with 3.12 binaries.

I can't help you more than that since I don't know what you're trying to run.

[wet surge]

The stable diffusion XL https://github.com/Stability-AI/generative-models/blob/main/README.md

[kindred tinsel]

it hasn't movedf in a while from

[vocal burrow]

Unless you're a dev (I doubt it) you should probably not use the original / official repository. It's meant for devs and scientists.
Use something like auto1111's stable-diffusion-webui, comfyUI, StableSwarmUI, etc

[vocal burrow]

makre sure that your cmd still has the focus (click on it and then press space or something)

[karmic crown]

if you got an error in terminal (cmd), it will never move

[vocal burrow]

not necessarily

[kindred tinsel]

It just said that some preprocessors wouldn't work and kept going twice for now

UNLES

[vocal burrow]

you can also sends what's already in it.

[kindred tinsel]

That error code surprises me, seems as if it were an exit code of some sort

[tame yoke]

masterpiece, best quality, masterpiece, asuka langley sitting cross legged on a chair
Negative prompt: lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, normal quality, jpeg artifacts,signature, watermark, username, blurry, artist name
Steps: 28, Sampler: Euler, CFG scale: 12, Seed: 2870305590,

[karmic crown]

as i said it will never move

[kindred tinsel]

@vocal burrow

Here is the log up to this point, all in the end are the errors

The webui is not dead though,

I can generate images just fine, I am guessing controlnet wouldn't work at all

[karmic crown]

Command: "C:\StableDiffusionIDES\stable-diffusion-webui-directml\venv\Scripts\python.exe" -m pip install mediapipe --prefer-binary
Error code: 1

[kindred tinsel]

yes

[vocal burrow]

close it, remove the extension (delete its folder in extensions+delete venv folder), run pip cache purge and try to install it again

[kindred tinsel]

are you fr? :cccccccccc install all of that again?!

[karmic crown]

not all

just venv

[kindred tinsel]

the good thing is that the sd drive moves wayyy faster now

[vocal burrow]

I didn't say install all of it again.

[kindred tinsel]

Did this, or do you want a system wide cache purge?

[karmic crown]

Collecting torch==2.0.0
  Using cached torch-2.0.0-cp310-cp310-win_amd64.whl (172.3 MB)

some of those cached files make problem

[kindred tinsel]

I mean all of the venv again, since it takes a while for all of it

[vocal burrow]

good

[kindred tinsel]

Running webui again :=