🤝｜tech-support

[languid tinsel]

https://tenor.com/view/mass-effect-repete-commander-shepard-shepard-you-what-gif-13879410

hwat

[cursive cobalt]

#🤝｜tech-support message 😄

[ornate elk]

Ohh no :c

Which tutorial did you followed

[cursive cobalt]

Please Stand by.

https://www.youtube.com/watch?v=kqXpAKVQDNU this guy

[languid tinsel]

it can't be bad, says it's #1 right there on the thumbnail

[cursive cobalt]

I mean. It did help me get it working, where a written guide did not

[ornate elk]

Remove ever argument with -- there.
Then only add:
--xformers --medvram-sdxl --no-half-vae --autolaunch

[cursive cobalt]

Wait... sdxl? I need the sdxl pack for those promts?

[ornate elk]

No

[cursive cobalt]

And was it half vae?

[ornate elk]

Its for compatibility

[languid tinsel]

it's a fix to a stupid issue that was introduced a few months ago xd

[cursive cobalt]

So... Its ready to be saved, once current batch is done, but what exactly do they do?

[ornate elk]

These let you generate images at the fastest speed and the lowest vram resources possible

[cursive cobalt]

So no more waiting 2 hrs?

[ornate elk]

The stuff you had isnt needed for a 1080ti and only slows down everything

[languid tinsel]

also for the love of god look into img2img, don't just do text2img xD

[cursive cobalt]

I guess thats what u get for not being bale to afford the newest gpu's each year xD

[ornate elk]

You will be amazed how fast it is now xD

[cursive cobalt]

Its mainbly used to change the style of an existing img, ye?

[ornate elk]

Na I've used a 1080 a whole year with SD. It can do mostly everything

[cursive cobalt]

1080 ti. GOTD.

Gpu of the decade.

[languid tinsel]

txt2img is what you use to generate a low resouliton image you start with, and then you refine it with img2img and upscale it to higher resolutions.

you shouldn't really be using txt2img for anything larger than 1000x1000 px, it creates issues

[cursive cobalt]

Wait... So txt2img = Prototype image. Img2img = production Image? (Using terms I am familiar with)

[languid tinsel]

yeah pretty much

[ornate elk]

Thats not true for sdxl and hires fix. But for 1.5 models

[cursive cobalt]

So i should only do it for 1.5 stuff, but not sdxl/pony etc?

[languid tinsel]

yeah but what's the point of generating 10000 pitures hoping to score gold instead of generating 100 and just refining the best one to be gold

[ornate elk]

No if you edited the webui-user.bat like I said you can use sdxl and pony and it will be fast

[cursive cobalt]

I c

Mmmm slow

[ornate elk]

Depends on the GPU ^^ but yea better to gen lowres and upscale only the good ones

[ornate elk]

What are your settings ?

[ornate elk]

If you enabled hires fix then always set the hires steps to 10

[ornate elk]

Don't use 68 steps. 25-35 is mostly enough

[cursive cobalt]

i c

[languid tinsel]

I'd set the initial size to 640x960, it's then easier to upscale to 1280x1920 and 2560x3840 which are pretty good resolutions to work with

[cursive cobalt]

And then swap them, should a landscape be needed?

[languid tinsel]

yup

[fair oxide]

Guys amd 7900xtx or rtx 4080 for Sd?

[cursive cobalt]

Can those be saved somewhere, so they're default? Also 4080. Nvidia just works better

All i saw, while trying to find resources to install this, was amd needed extra files

[ornate elk]

I'm using a 7900XTX

With SD. But I bought it for Gaming as its better than the 4080

[languid tinsel]

amd has some hiccups but there are workarounds when you know what you're doing

[fair oxide]

amd has some lacking right they cant use hip so they have to install 3d part software for ad to use their hip which is similar to cuda

[ornate elk]

If you want it just for AI stuff then go for nvidia as the support is better

[fair oxide]

what is it called

[fair oxide]

how many steps can i go with 4080

[cursive cobalt]

@ornate elk Halp.
Wat dis mean? Came up after changing the promts, and now the Lora tab is infinity loading.

[ornate elk]

No we have hip SDK and rocm support but pytorch doesn't support it on windows right now.
So on windows I use Zluda to emulate cuda

Getting 16 it/s

[cursive cobalt]

Not promts** sorry. The thingy

The... Ehm........ --stuff

[fair oxide]

yeah i wanted to say zulda

[ornate elk]

?

[fair oxide]

any updates on amd doing anything for sd for better support?

[ornate elk]

@cursive cobalt What's in Your webui-user.bat rn

[fair oxide]

to use HIP instead of zulda

[fair oxide]

its been like years what since people using zulda for amd

[ornate elk]

Yes they still work on it. Rocm development goes further and we hope to get windows support this year

[ornate elk]

Its since February ... XD

Before we only had directml

[fair oxide]

wow!so should i go with amd 7900xtx if support coming this year?

that vram wil work better than 4080 right

do 7900xtx and 4080 perform same in sd

[ornate elk]

I can't guarantee it. But if like I said if you want to do much ai stuff your safer with nvidia.
I bought my 7900xtx knowing that I can use it not as fast as with nvidia cards on windows. But now with Zluda I'm happy until full rocm support comes

[fair oxide]

have u used videos to convert into anime in SD?

i mainly want to use sd for that

or wait for 5080

this year

i want to convert 1 minute video of 1080p@60fps @30fps into anime in SD

how much time it will take with 4080

[ornate elk]

Like video onto frames, then using Controlnet and then put it back together as video?
Then yes I've done that

[fair oxide]

or 4080 super

[fair oxide]

how much time it took u

[ornate elk]

Looks good. Try restart and then check the lora tab again

[ornate elk]

Depends on the length idk rn

[fair oxide]

1 minute video

1080p@30fps

to anime

how much time with 4080

or 7900xtx

[cursive cobalt]

... I might need to move all of it so a quicker drive...

Can i just copy paste the main folder, pythin, git and SD (And Auto) is installed in, and just launch?

[ornate elk]

If you can send me a video I'll try it later

[fair oxide]

okays thx bud

can u dm me?

[ornate elk]

If you move the webui it will break the auto update (git path)

But its fixable

[fair oxide]

1 of the user said here i forgot his name that to convert 1 minute video into anime it will take 3 hours by some math he explained

is 7900xtx good for video editing?

[cursive cobalt]

Fairly ez, or bothersome?

It takes me like a solid 5-10 min launching the webui xD

[fair oxide]

so math he explained was 60 secs into 60fps

and do that into hours

anyways u tell me bud

how much time it takes for u @ornate elk

ill send u video here or dm?

[ornate elk]

In dm

I can test it later when I'm at home

[ornate elk]

Fairly easy

But don't move git and python

Only the stable-diffusion-webui folder

[cursive cobalt]

Once I've cleared up the space needed for my needs, could you guide me step by step?

[ornate elk]

You mean to fix the auto update stuff?

[cursive cobalt]

Is that the only thing thats gonna get broken?

[ornate elk]

Yes

The webui will work Normaly

[cursive cobalt]

Will it speed up the loading, moving the sd folder from hdd to ssd?

Wait... That is a silly question. Obviously it is, if the things its pulling `such as Lora's, is on the same drive

[ornate elk]

Yes

It won't affect image generation time

[cursive cobalt]

Cuz thats gpu

Cause spending 800+ seconds on loading all of my Lora's, is a bit... excessive xD

OH BOI! Thius is gonna take some time... output disc is hovering at ~40 MB/s

ETA: 2½ hrs

[valid gorge]

Trying to use Pixart Sigma, but getting this error. Ideas how to fix it? I should have all the filex downloaded and placed into correct folders. Trying to use the "Abominable Spaghetti Workflow - PixArt Sigma" workflow from Civitai.

[cursive cobalt]

I have been using the same drives since at least 2014, @ornate elk ... Except C drive.

Been using that since 2018 xD

[ornate elk]

Quite full xD

Make sure C has more free space

[cursive cobalt]

It doesn't help, that currency the drive of which i am trying to install this on (the Evo 870), has over 100 gb of unaccounted missing space

And i checked partitions

[ornate elk]

How does it look in the partition manager?

[cursive cobalt]

Everything is partitioned correctly, so idfk.

I QUESS that's what happens with 10 yrs of unaccounted random stuff on your drives, does to it

Speed*

[ornate elk]

If the 870 is a 250gb SSD its normal

Or where you miss space?

[cursive cobalt]

I do be missing the space on the 870,but the output (or where I fra copying from) is an old hdd

But is fix now. We good

[pulsar vessel]

I have the RXT 6800, I’ll look at your pinned messages

[sonic sphinx]

why cant i join mid journey server

[solemn atlas]

Hii, what does that mean?

[scarlet crater]

what happened to mask_mode in api?

[hushed turtle]

I logged in VIA my Discord on Stable Diffusion. Now the loggin method isn't there anymore to do the API credits.

[uncut jewel]

Anyone can help on that issue by any chance ? 🙂

[uncut jewel]

I'm not 100% sure but it might mean that you use 2 different models that arent compatible (for example an sdxl checkpoint model with a sd1.5 controlnet model)

[cursive cobalt]

Hey @ornate elk fren.

Do you happen to know, whether its possible to bind the webui.bat file, to something, so I don't need to navigate to the folder each time, to start up SD?

[ornate elk]

Then follow my AMD Zluda guide and you will have a fast stable diffusion programm 🙂

[pulsar vessel]

Should I go through and uninstall everything I have already done?

[ornate elk]

Make sure you always start the webui-user.bat.
Make a right click on it and press Send to > Desktop
Then you have a shortcut

[cursive cobalt]

Oh nice

Is there a command to free up the VRam after each generated batch? Having the webui crash on me a few times cause it claims it doesn't have enough vram

[ornate elk]

If you already have git and python you don't need to uninstall them but delete the old stable-diffusion-webui folder

[cursive cobalt]

Wanted to bind it to my taskbar xD

[ornate elk]

Make sure to use batch count and not batch size

[cursive cobalt]

Wait... Thats whats consuming a shit ton of vram?`Oh

I assume it frees it back up quicker as well then..?

Tis' causing issues on new gen

[scarlet crater]

Have you figured? I got this after the update.

[ivory skiff]

Hello, anything i try to generate it generates badly

and takes ages

[ornate elk]

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

[ivory skiff]

i have been thru 2 gpus

the 1060 6gb and now 3060ti

and it s the same

it is a fresh install of the program

it used to work some time ago

[cursive cobalt]

What could be causing this issue? Its essentially crashing during generation

[ornate elk]

You need --xformers --medvram-sdxl --no-half-vae in your webui-user.bat

For the best performance

At the line commandline_args

[ivory skiff]

so xformers being disabled generates complete jibberish?

is this a problem by any chance?

it is exactly the same...

1.5it/s untill the 50% mark, and then extremely slow!

and still random things

even with comfy ui it does the same

[cursive cobalt]

..... Excuse me wtf... Started getting grey squares genning... I look through my checkpoints... They're all gone. Wat

[ivory skiff]

daamn

[thorny valley]

Hey guys, anyone know how to solve this error when booting a1111?
ImportError: cannot import name 'Undefined' from 'pydantic.fields'

ive been using a1111 every day for a very long time now and today when running it it wasn't able to generate anything so i ran "git pull" and now im getting that error

[stone hawk]

Hello! I am currently trying to train a character lora with https://colab.research.google.com/github/hollowstrawberry/kohya-colab/blob/main/Lora_Trainer_XL.ipynb#scrollTo=OglZzI_ujZq- for use with PonyXL.

Everything seems to load fine, it picks up my dataset, all going well. Then, after about 5 minutes, I get "torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 480.00 MiB. GPU 0 has a total capacity of 14.75 GiB of which 307.06 MiB is free. Process 60818 has 14.45 GiB memory in use. Of the allocated memory 14.01 GiB is allocated by PyTorch, and 314.76 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables)"

I tried running the line it suggested, but got a syntax error. Where do I go from here? I would appreciate any help, thanks!

[fair oxide]

@ornate elk u finished with the video?or is it still processing?

[ornate elk]

What's your txt2img settings?

And model?

[ornate elk]

I was at work till now xD

[fair oxide]

ooh

so can u do now the 1 minute video i sent u now or whenever ur free thx buddy

dm me ping me when u are finished with the video man

[fair oxide]

with the video or ur personal work?

[ivory skiff]

i have got an old image i made last year

put it into PNG info

then transfered all of that into the txt2img

with the same checkpoint

[lunar linden]

nope. had to switch to comfyui

[scarlet crater]

Oh alright I managed to do it

[lunar linden]

how?

[scarlet crater]

It seems I was setting latent_mask in the request which I shouldn't be

Removing it managed to fix it

[lunar linden]

ah true. tbh i prefer comfyui. really good how you can send an entire workflow through the api. give you a lot more control

[scarlet crater]

I may be can switch to it

Thanks for response

[lunar linden]

👍

[ornate elk]

Personal

[ornate elk]

Can you show the settings ?

Txt2img

[fair oxide]

ooh np

[ornate elk]

What's the model?

[ornate elk]

Also try DPM++ 2m as sampler

Or Euler A

[ivory skiff]

i was told karras was a good one

and i have been using it since

generated 400 pics with it

[ornate elk]

It is karras

Dpm++ 2m will use karras

[ivory skiff]

oh, okay

and why do everything i do

generate that stuff

blue image

even with confy ui

[ornate elk]

Because the model may be corrupt or your using the wrong vae

Have you tried other models?

[ivory skiff]

1 sec my pc is dying from wsl using 16gb of ram

I will download another model

[fair oxide]

https://rightbrainai-1307270943.cos.accelerate.myqcloud.com/2024-05-17/22-02-ee2c1dae-7538-4634-883f-f944ef36c609.mp4

thats me guys!wow anime is so cool!

[craggy patrol]

What am I doing wrong with the api?

{
    "session_id": "AB3A76D76053E20A6933A2B60797DA8E31553410",
    "images": 1,
    "rawInput": {
        "prompt": "a photo of a cat",
        "model": "cosxl",
        "steps": 20,
        "width": "1024",
        "height": "1024"
    }
}

to http://192.168.0.14:7801/API/GenerateText2Image

Generate:

13:48:22.835 [Warning] T2I image request from user local had request parameter 'rawInput', but that parameter is unrecognized, skipping...
13:48:22.835 [Info] User local requested 1 image with model ''...
13:48:22.837 [Error] Internal error processing T2I request: System.NullReferenceException: Object reference not set to an instance of an object.
   at StableSwarmUI.Builtin_ComfyUIBackend.WorkflowGenerator.CreateStandardModelLoader(T2IModel model, String type, String id, Boolean noCascadeFix) in /src/BuiltinExtensions/ComfyUIBackend/WorkflowGenerator.cs:line 1348
   at StableSwarmUI.Builtin_ComfyUIBackend.WorkflowGenerator.<>c.<.cctor>b__10_0(WorkflowGenerator g) in /src/BuiltinExtensions/ComfyUIBackend/WorkflowGenerator.cs:line 90
   at StableSwarmUI.Builtin_ComfyUIBackend.WorkflowGenerator.Generate() in /src/BuiltinExtensions/ComfyUIBackend/WorkflowGenerator.cs:line 1308
   at StableSwarmUI.Builtin_ComfyUIBackend.ComfyUIAPIAbstractBackend.CreateWorkflow(T2IParamInput user_input, Func`2 initImageFixer, String ModelFolderFormat, HashSet`1 features) in /src/BuiltinExtensions/ComfyUIBackend/ComfyUIAPIAbstractBackend.cs:line 627
   at StableSwarmUI.Builtin_ComfyUIBackend.ComfyUIAPIAbstractBackend.GenerateLive(T2IParamInput user_input, String batchId, Action`1 takeOutput) in /src/BuiltinExtensions/ComfyUIBackend/ComfyUIAPIAbstractBackend.cs:line 701
   at StableSwarmUI.Text2Image.T2IEngine.CreateImageTask(T2IParamInput user_input, String batchId, GenClaim claim, Action`1 output, Action`1 setError, Boolean isWS, Single backendTimeoutMin, Action`2 saveImages, Boolean canCallTools) in /src/Text2Image/T2IEngine.cs:line 255

[violet swift]

any solution for this?

File "C:\Users\kevin\Desktop\ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-AnimateDiff-Evolved\animatediff\motion_module_ad.py", line 1112, in forward
x = x + self.pe[:, : x.size(1)]
^~~~~~~~~~~~~~~~~~~~~~~
RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!

[ornate elk]

Yea I know how to fix it. I'll send you a guide when I'm at the PC

[ivory skiff]

try disabling the.. cpu rendering?

[stone hawk]

Having trouble while training a lora with google colab, can somebody who know's their stuff dm me?

[violet swift]

how do I do that? I run the GPU

[ivory skiff]

Colab pro?

[ivory skiff]

I don t know your setup

[violet swift]

i run this,

[ivory skiff]

And what is inside of that

[ivory skiff]

Can you try to use confyui's ui?

[violet swift]

ye

I have deleted everything and I am reinstalling

[fair oxide]

guys is StoryDiffusion part of stable diffusion?apparently it can convert 30 secs video into more anime looking quality consistent anime looking quality which we see in anime so any1 heard of it?

[thorny valley]

Thank you!

[ornate elk]

that pydantic error appears when using a not updated controlnet version.
Can be resolved by doing this:
Go into the stable-diffusion-webui folder. Click in the explorer bar (not search bar)
Type in cmd and hit enter.

Run the three commands one by one:

venv\Scripts\activate pip install albumentations==1.4.3 pip install pydantic==1.10.15

Then relaunch the webui-user.bat

[stone hawk]

No idea, I'm using this one: https://colab.research.google.com/github/hollowstrawberry/kohya-colab/blob/main/Lora_Trainer_XL.ipynb#scrollTo=OglZzI_ujZq-

[ivory skiff]

and you can train ai on google colab?

they banned ai generation last year

training works still?

[stone hawk]

I guess not, since mine isn't working

[prisma jackal]

Hi, I have an AMD 7900 XT and I followed the tutorial pinned for AUTOMATIC1111 with ZLUDA.
I would like to know if there are any optimizations in COMMANDLINE_ARGS that I can add to reduce VRAM consumption since when I generate with SDXL with hires fix it consumes all the available VRAM.
Now I'm using these arguments: set COMMANDLINE_ARGS=--use-zluda --medvram-sdxl --update-check --skip-ort, thanks.

[thorny valley]

hey sorry to bug you again, but i actually ended up doing a clean reinstall of A1111 and i tested it w/ one image, ran fine. but now while the UI boots, it won't generate anything and i dont even get any errors...

gets stuck like this:

[ornate elk]

Hey I have the 7900xtx and what you can do is to install the tiled vae extension to reduce the vram usage at upscaling

Also use the sdxl fp16 vae

[ornate elk]

which install guide did you followed?

[thorny valley]

the one on the main github page. i have a nvidia gpu (4070ti)

[ornate elk]

i would recommend installing it using my install guide

Here: #🤝｜tech-support message

[thorny valley]

trying that now :)

[violet swift]

final error, hel pls

[thorny valley]

still reacting the same way after re-installing: (now it gets stuck loading the model with the thinking icon)

[ornate elk]

make sure the webui is whitelisted in any browser adblocker

also try loading in an other model

[thorny valley]

i have no adblockers

[ornate elk]

okay

are the models on hdd or ssd?

and the webui

[thorny valley]

nvme ssd where my os is installed

i dropped dreamshaper8 model into the folder and ran the ui. while it was stuck thinking i try to switch over and it lists the new model but its just stuck thinking with no error in cmd prompt

you dont think this error is anything i need to address right? "FutureWarning: resume_download is deprecated and will be removed in version 1.0.0."

[ornate elk]

that can be ignored

[ornate elk]

how much ram do you have?

[thorny valley]

32 gb

[ornate elk]

also make sure to not run any other heavy programms or wallpaper engine in the background

[thorny valley]

hmm... the thinking stopped and gave me an error. i refreshed model list, selected dreamshaper and was able to run an image

lemme see if I can get consistant restarts of the ui and generate things

what a bizzare series of errors ive never had before... been using a1111 for almost 2 years now

[ornate elk]

very strange indeed

[thorny valley]

reloaded ui and it locked up again...

[ornate elk]

rip

maybe restart the PC

also make sure the gpu driver is updated

[languid tinsel]

I'm redoing some preview images for checkpoints and suddenly one of them started doing this. Weirdly they look completely fine during generation, it's something that happens as the pictures save. ideas?

[thorny valley]

happens to me when I forget to turn off my vae

[languid tinsel]

I'll try without vae

yep that was it. Odd. Thanks

[ornate elk]

wrong vae mostly

[thorny valley]

rebooted pc, same issue. i randomly thought to try on edge browser instead of chrome and it worked again. gonna keep testing a bit and see

ty so much for your time btw. been pulling my hair out w/ this today lol. was working perfectly every day until last night

[ornate elk]

no problem

what changed since yesterday? any update?

[thorny valley]

i guarantee it was an issue with chrome. which caused me to delete the venv folder to reset it which is when that pydantic error popped up

which led to the a1111 reinstall and more issues w/ chrome

so far everything's working on edge

wish i had tried that before reinstalling it all and having to re-do everything lmao

[ornate elk]

true, but know you have a fresh webui

thats good too ^^

[violet swift]

hello, solution for this?

File "C:\Users\kevin\Desktop\ComfyUI\ComfyUI\custom_nodes\ComfyUI-AnimateDiff-Evolved\animatediff\motion_module_ad.py", line 1112, in forward
x = x + self.pe[:, : x.size(1)]
^~~~~~~~~~~~~~~~~~~~~~~
RuntimeError: Expected all tensors to be on the same device, but found at least two devices, cuda:0 and cpu!

[pulsar vessel]

@ornate elk I get to here in my command prompt "Launching Web UI with arguments: --use-zluda --medvram-sdxl --update-check --skip-ort" and nothing happens.

I follwed your ZLUDA guide, any thoughts?

wait...it took 2 min to load...ignore^^^

[ornate elk]

so it works now? ^^

[pulsar vessel]

SD opened in my browser, Im trying my first image generation with the knowledge that it could take 15-40 minutes. I will update if I make progress. thanks!

[ornate elk]

perfect 🙂 if the generate button is grayed out after pressing and in the cmd is showing no error, then let it load and you get an image

[pulsar vessel]

sounds good, ill be patient, thanks!

[ornate elk]

np, mostly its faster than 20 minutes

[solemn atlas]

happend to me too today

[pulsar vessel]

we are all set, I think i saw somewhere that you can't train with AMD, is that true?

[ornate elk]

yes thats true

its only possible with the OneTrainer tool, because it supports zluda

[pulsar vessel]

good to know, thank you

[ornate elk]

no problem 🙂 wahst your gpu btw?

[pulsar vessel]

RXT 6800

or RX 6800 XT maybe

[ornate elk]

ah nice

great card

[pulsar vessel]

great for gaming 😄

[ornate elk]

for sd it will be good too, not as fast as nvidia but you can even use sdxl models and upscaling without problems

[pulsar vessel]

awesome, looking forward to playing more with it, thanks!

[ornate elk]

no problem 🙂 for any questions feel free to ask

[pulsar vessel]

will do, appreciate all the help. I will say, im not sure if it was just me, but the instructions on this site https://www.wikihow.com/Change-the-PATH-Environment-Variable-on-Windows were a little unclear because wherever it told me to click it a button, it was only showing special characters

I ended up getting it, but it was a little annoying, not sure if you want to find a new one

[ornate elk]

for me on firefox it looks like this:

so maybe a browser issue or browser extension issue

[pulsar vessel]

possible, good to know though

[violet swift]

⚰️

[ornate elk]

What's your GPU?

[violet swift]

1660 GTX 6GB

[ornate elk]

Okay, make sure your nvidia driver is updated

[violet swift]

ok i go to check

[fallow shoal]

Why does it take so long time to "moving model(s)"?
Have been around 1h and 30mins and still only at 50%
Using a 2060 12gb

[violet swift]

updated, do I try running the program?

[ornate elk]

Yep

[ornate elk]

Where do you move them?

[fallow shoal]

I am not really moving them, I just downloaded a model and put it in the models folder and tried to create a picture with txt 2 img but it said in the cmd that it was moving model(s)

But I read https://www.reddit.com/r/StableDiffusion/comments/15qwvo8/a1111_model_taking_forever_to_load_and_freezes_my/
That RAM speeds up the process a lot

[ornate elk]

For sdxl models you need 16gb ram or more

If you only have 16gb you can increase the windows pagefile

[violet swift]

same error

[thorny valley]

@ornate elk hey everything's running buttery smooth. one last quick question: after I install the TensorRT extension can I just drop the onnx models I backed up? or do I need to recreate them?

[fallow shoal]

I have 16gb but I will try increasing it, thanks for the help ❤️

[ornate elk]

Wouldn't recommend using tensorRT as its not compatible with everything SD related.
But yes you can drop the models back in

[thorny valley]

hmm ok. been using it for a long time for non sdxl models w/ no issues. anything i should be concerned with?

[ornate elk]

No if it works then its okay, but it could be that some extension won't work with it

[thorny valley]

i recently upgraded from a 3060 w/ 16gb ram and back then had a hell of a time running sdxl models

[thorny valley]

gotcha. ty :)

it doesnt work w/ xl models or controlnet from my experience

[ornate elk]

Ah alright that's what I thought

Sdxl should work fine for 3060 12gb

[thorny valley]

but does wonders w/ 1.5 models and now its integrated w/ a1111 wonderfully and no longer have to embed loras to each individual onnx model

[half island]

Hoi, is there a way in python code/comfyui/forge to make SD ignore steam deck's bios UMA frame buffer limits? As i am trying to generate images on the steam deck, but oddly comfy won't do anything past 4GB that i set in uma.

[torn salmon]

Can anyone help me w my account?

I paid for the monthly and there is no button that says Discord activation so I can use the bot here???

[half island]

They will be in a text file that tells the extensions that "these are trained, use those", thus list them. So you can move them away as you wish. Only "recreate models" you need to do is per resolution/resolution range, as it still doesn't have a effective way to make the model entirely TRT

[orchid temple]

hey everyone, this is SD 1.5, right?

this is in the models folder of my Automatic1111 SD WebUI install

I checked, it is 1.5

how do I upgrade to SDXL?

[primal axle]

i am having trouble setting up my stable diffusion checkpoint stuff, msg me i am like a newb when it comes to all of this. someone plz help

[onyx solar]

SDXL is not an "upgrade" in the traditional sense. Automatic1111 is the software. SD 1.5, SDXL, and others are different models. Automatic1111 can run both with no changes to the software you have installed. Simply place the SDXL models in the same place and ensure that you download the SDXL VAE and put it in its proper folder. You will then be able to produce images using either model. (I recommend using subfolders for each model type within your model folders.)

[orchid temple]

how is SDXL not an upgrade over SD 1.5?

[thorny valley]

thank you!

[onyx solar]

It is an upgrade in the sense of being a more capable model overall. It is not an upgrade in the sense of needing to change your software.

That's why I said it wasn't an upgrade in the traditional sense.

[orchid temple]

Yeah I knew that, I was just unsure if it was just as simple as downloading the SDXL .safetensors file and putting it in here

[onyx solar]

And I answered that.

[orchid temple]

yup

ty

[onyx solar]

Not a great way to get help here. You should share more specific information about what you are running into. Any errors, screenshots of examples, etc. will be useful for people to answer/guide you.

[half island]

A model upgrade would more so be for instance SDXL models that doesn't need refiners, as they are finetuned to be fine with just one model. That's more a upgrade. SDXL is just it's own model capable of higher resolution. But even today, i find sdxl a "downgrade" over the better 1.5 models that can do so much more.

[orchid temple]

my 1.5 model doesn't get me very good images

I assumed SDXL would also have better image quality

I don't really care about resolution

how do I find better SD 1.5 models?

I'd prefer that to downloading SDXL which is like 7 gb

and then the VAE is another 7 gb

[half island]

Show a result you've made. make a random dog/car/cat one for instance.

[orchid temple]

yeah one sec

[half island]

As SD 1.5 can look quite good once you've stepped it up, as well as upscaled it

[ornate elk]

your using the default 1.5 model its just bad. you need to download community made models from civitai.com like dreamshaper v8

[half island]

Oh yeah, default any model is complete ass by itself

[orchid temple]

ah

that makes sense

let me download something from civitai then

[half island]

Upload your gen, and i'll remake it in dreamshaper as an example of before and after :P

[orchid temple]

too late

I don't like to generate images like these though

I usually do like

architecture or technology

[half island]

Any will do, i just threw out examples :P

[orchid temple]

almost never humans or animals

[half island]

Browse the output folder and drop in one of your liking/attempts

[orchid temple]

I made some space elevators a couple days ago

I think these were the best I had

[half island]

Here's those results in a custom SDXL

[orchid temple]

oh my

link?

@half island

[half island]

Then we got dreamshaper 8 as mentioned before ;P Using prompts of the second image.

Model of the 2 above was with https://civitai.com/models/134338/sdxl-faetastic-details

https://civitai.com/models/4384?modelVersionId=128713 Here's for dreamshaper v8

[viscid gate]

webui not opening after restart

[ornate elk]

its not installed correctly, and also the wrong python version.
please follow my install guide:
#🤝｜tech-support message

uninstall python 3.11 before starting the guide

[viscid gate]

it was just working until today

i have a different issue now

i just get this big ass error message now webui is open

ive followed all ur steps

it says out of memory and all im so confused

and my memory usage is at 0

[ornate elk]

No you didn't. In the screenshot it still says stable-diffusion-webui-master and not stable-diffusion-webui.
You also didn't edited the webui-user.bat.
Maybe you just launched your old install again instead of the new one?

[ornate elk]

It means GPU vram

[violet swift]

i have 2 errors, help pls

[timid notch]

hello, i wanna reinstall comfyui, so should i delete the whole package?

or maybe it has some sittings can make the software to default ?

[spice gazelle]

if i installed new ram today would that cause the program to have to load for longer before use? or was it the new update have it set to git pull auto update upon opening the webui-user.bat cmd said everything loaded propberly but the ui was frozen/loading for 15 minutes before it decided it would work?

[formal trout]

What would be a good parameter for inpainting/soft-inpainting so I don't get that halo? I'm trying to fix some upscalling seems.

working on Automatic1111

[cursive cobalt]

Just taken a look at this. Intended use: Making sure each individual in a scene, gets the correct hair color.

https://github.com/hako-mikan/sd-webui-regional-prompter

Anyone used it?

[formal trout]

I used it a little a while ago, with DS 1.5, it is good to try to make some kind of composition

But it is very evasive, it may do what you want or not

I used again to get different characters and it was completly frustrating

I mean your prompt "A boy AND a girl" and it does a random character in the middle, something like that

[vapid sluice]

Is there a way to only reinstall the run.bat file for comfyui ?

[vapid sluice]

I get a random unrelated software in my computer loading stuff midway through the start up log, and the log just stops for a moment until i close the pop up window of said software

[vapid sluice]

Or should i reinstall comfyui 💀

[cursive cobalt]

The kind folks in Anime helped me :3

[formal trout]

Oh I'm reading 🥸

[formal trout]

But see, it seem to only work on changing hair on the 3 same duplicated girls 😂

[cursive cobalt]

I'm getting it to work, just learning the basic settings tbh

[formal trout]

it works better to establishing features for one single character, but multiple different characters, it cannot handle, maybe after a lot of tries

[copper hemlock]

hello, I was hoping someone could help me with an issue I'm having, I cannot get Stable Diffusion to generate anything, everytime I try, I get the same error message

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

Stable diffusion model failed to load
*** Error completing request
*** Arguments: ('task(z84q6hy74ibixlo)', <gradio.routes.Request object at 0x000002827142A200>, 'boy', '', [], 1, 1, 7, 512, 512, False, 0.7, 2, 'Latent', 0, 0, 0, 'Use same checkpoint', 'Use same sampler', 'Use same scheduler', '', '', [], 0, 20, 'DPM++ 2M', 'Automatic', False, '', 0.8, -1, False, -1, 0, 0, 0, False, False, 'positive', 'comma', 0, False, False, 'start', '', 1, '', [], 0, '', [], 0, '', [], True, False, False, False, False, False, False, 0, False) {}
Traceback (most recent call last):
File "C:\Stable Diffusion\stable-diffusion-webui-master\modules\call_queue.py", line 57, in f
res = list(func(*args, **kwargs))
File "C:\Stable Diffusion\stable-diffusion-webui-master\modules\call_queue.py", line 36, in f
res = func(*args, **kwargs)
File "C:\Stable Diffusion\stable-diffusion-webui-master\modules\txt2img.py", line 109, in txt2img
processed = processing.process_images(p)
File "C:\Stable Diffusion\stable-diffusion-webui-master\modules\processing.py", line 832, in process_images
sd_models.reload_model_weights()
File "C:\Stable Diffusion\stable-diffusion-webui-master\modules\sd_models.py", line 860, in reload_model_weights
sd_model = reuse_model_from_already_loaded(sd_model, checkpoint_info, timer)
File "C:\Stable Diffusion\stable-diffusion-webui-master\modules\sd_models.py", line 793, in reuse_model_from_already_loaded
send_model_to_cpu(sd_model)
File "C:\Stable Diffusion\stable-diffusion-webui-master\modules\sd_models.py", line 662, in send_model_to_cpu
if m.lowvram:
AttributeError: 'NoneType' object has no attribute 'lowvram'

If anybody knows how to fix this I would greatly appreciate the advice

[fallow ginkgo]

hi guys i need help, i have generated a very good emoji and i want to make more then 1 of it but with different face and so on can someone tell how?

[fair oxide]

Anyone facing xformers error while using --xformers?

Always when i try generate pic it fails on 90% and shows xformers and CUDA like error

Same thing while using sdxl safetensors

[ornate elk]

Hey, you need to install the webui the right way using my guide from the pinned messages of this channel.

[ornate elk]

What error do you get?

In the cmd

[ornate elk]

Okay can you show the whole cmd log please

Also what's your GPU

[fair oxide]

Nvidia RTX 2050

[fair oxide]

alright. let me take screenshot

[meager frigate]

Will sd3 available on replicate?

[ornate elk]

No one knows

We need to wait for the public release

[meager frigate]

Ok thanks, Hope that will be soon 🥲

[ornate elk]

Me too 🥲

[fair oxide]

Yup

[ornate elk]

Ah you installed it by downloadind the zip file?

[fair oxide]

yes correct

it worked fine yeasterday but today i got getting these errors

[ornate elk]

Then first run the Update.bat
Then delete the venv folder
Then edit the webui-user.bat and there add --xformers --medvram --no-half-vae
Then save and relaunch

[fair oxide]

alright i`ll try that one moment

and then... here is the another issue. i not find venv folder

bcs its not there

nvm found it and followed all steps but still xformers WARNING and errors.

WARNING:xformers:WARNING[XFORMERS]: xFormers can't load C++/CUDA extensions. xFormers was built for:
PyTorch 2.1.2+cu121 with CUDA 1201 (you have 1.13.1+cu117)
Python 3.10.11 (you have 3.10.6)
Please reinstall xformers
Memory-efficient attention, SwiGLU, sparse and more won't be available.
Set XFORMERS_MORE_DETAILS=1 for more details
No SDP backend available, likely because you are running in pytorch versions < 2.0. In fact, you are using PyTorch 1.13.1+cu117. You might want to consider upgrading.

[ornate elk]

What folders do you have?

[cursive cobalt]

What is "Clip skip" exactly, and why do i see a fair few items on civitai having them in the details section?

[fair oxide]

without --xformers it works fine with default sd models but still fails on sdxl models on 90%

[fair oxide]

3050 sorry typo

[ornate elk]

hm okay, i would then recommend to install the webui using my install guide.
then you get an venv where its easier to fix stuff

Here is the guide:
#🤝｜tech-support message

[fair oxide]

git config --global --add safe.directory F:/Ai/stable-diffusion-webui
fatal: detected dubious ownership in repository at 'F:/Ai/stable-diffusion-webui'
'F:/Ai/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 F:/Ai/stable-diffusion-webui

[ornate elk]

that means your drive is not formated correctly

[fair oxide]

ah so i need deep clean format?

[ornate elk]

yep, it needs to be gpt formated in the diskmanager and then ntfs

[fair oxide]

alright

[ornate elk]

but save your stuff before formating ^^

[fair oxide]

okay bro 🙂

[fair oxide]

works now. thx ❤️

[ornate elk]

perfect, np 🙂

[heavy perch]

@ornate elk need your help again 😅

How do I fix WinError2?

[azure sand]

I tried sd 8 months ago. When the new game driver nivida wasn't working properly and it only working with 531.5 version. Did that glitch fixed in newer edition. And is there any other method created which can create result faster? Or anything new in that regard

[ornate elk]

Would need to see the complete cmd log

[ornate elk]

Its fixed because most people use the latest driver without issues

[azure sand]

Ok

[fair oxide]

i have to say... ur guide is much better than one-click install

[ornate elk]

Thanks 🙂

i tried to cover all steps for the best way to install it

[placid pier]

Hey guys, does anyone know a way to generate without meta data being saved?

[ornate elk]

that isnt a complete cmd log

[open fox]

Are you talking about comfyui? You can startup the server with an argument --disable-metadata then no metadata is included in the image.

[ornate elk]

make sure the video file names dont contain special characters, also do you have ffmpeg properly installed and in path?

[placid pier]

No A1111
I forgot to mention

[cursive cobalt]

TROUBLESHOOTING TIME again @ornate elk 😄

[ornate elk]

what have you done before ?

installed any extension?

[cursive cobalt]

Please stand by. Might have fixed it

Ah yes. It had ommited to copy a specific file from when i was moving it, but i was smart, and copied it into another folder, before i formatted the drive. I found the files, and put them back in, and it launched

Can i add an image for this, so it doesn't show "No preview" but has an image instead?

[ornate elk]

yes thats possible in two ways

if you generate a image you can then click on the right corner of the lora or embedding, then press the edit icon and there click Replace Preview. then it will use the latest generated image

[cursive cobalt]

Oh really..? Nice

[ornate elk]

or you can manually place an image into the lora or embeddings folder and name it the same

[cursive cobalt]

Ah i c

Fanks

[heavy perch]

np for the names but not sure about ffmpeg, how can I check?

[ornate elk]

hmm if you run the install_bat it should have already made the ffmpeg path

[heavy perch]

yeah I did before

but still the same error

[mortal compass]

Some extensions also download pictures and info automatically

btw CS, I resolved my issues, thank you for the wise words!

[cursive cobalt]

GO on.... Auto1111 😄

[mortal compass]

For example https://github.com/BlafKing/sd-civitai-browser-plus
But keep an eye out after installing it, I think the extension may have upset some of my files or something. At least when I downloaded wildcards through it, then I got errors because it downloaded the pictures, json files and preview info

[cursive cobalt]

That doesn't bode well

[heavy perch]

@ornate elk what should I do

[solemn moon]

yuuussss
any1 skilled in QR Art?

[ornate elk]

you should open an issue on the roop-unleashed github, they will know where the problem is

[mortal compass]

Let's say I have an extension, and it downloads extra files upon being used the first time, and now I cannot find these files. Is there a command in terminal that can delete everything related to those files, so I don't need to role play as a detective?

[ornate elk]

you can open up a cmd and type
pip cache purge
and to fully delete those files you would need to delete the venv folder.
but after the first launch every file needed will be redownloaded

[mortal compass]

I'll try the purge

Last part sounds a bit scary lol

[ornate elk]

only scary if your internet connection is bad yD

[mortal compass]

I live in Sweden so should be no problem xd
Is venv in appdata? Its not in my forge folders lol

its not in appdata either wtf

suppose it's this bad boi

[ornate elk]

i thought forge has it too

[mortal compass]

I'll look it up

[blissful garnet]

Hey , coming acorss this problem.

It was running and installed all fine and dandy until I got a windows update, now I'm stuck with this issue.
I thought I had broke it playing around with comfyuis new fork, but even after a restore to previous state it still has the same error

[mortal compass]

webui-user.bat looks like this
`@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --xformers --autolaunch --pin-shared-memory --cuda-malloc --cuda-stream --api

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

call webui.bat`
But I don't have A1111 😳

[ornate elk]

looks normal

[mortal compass]

Ait well, I didn't find venv on their github so I guess I don't have it at all

[ornate elk]

it normaly gets created at the install

its not in a repo

[mortal compass]

the files im looking for is not in models either, this shouldn't be that hard right?? fuck

[ornate elk]

idk what files you are looking for

[mortal compass]

ops 😳 mb
C:\Stable_Diffusion\Forge\webui\extensions\sd-forge-layerdiffuse this extension downloaded a bunch of files before starting, then proceeded to not work and give me errors, and in the end not generate nothing. So I figured might as well delete it fully and try fresh

Shouldn't one of these create the venv folder? 🤔

[ornate elk]

forge has a venv folder for me

[mortal compass]

WHY THO

[cursive cobalt]

Where would .pt files go to? https://civitai.com/models/159990?modelVersionId=179964 That is a .pt file, when downloaded.
Auto1111 UI

[mortal compass]

How can it not install the venv for me when I installed it via the friggin one click installer

[ornate elk]

stable-diffusion-webui/embeddings

[cursive cobalt]

Thank u good sir

/madam,

/whatever you identify as 🙂

[ornate elk]

im an Ai

jk

[mortal compass]

found it C:\Stable_Diffusion\Forge\webui\models\layer_model xd

[mortal compass]

I'm kinda tired that forge is making me troubleshoot my webui every other day.. Thinking about installing the normal one like you suggested. I'm a little bit hesitant though, cause atm I'm generating images pretty fast, that's why most ppl get forge I suppose. Any idea if I will notice a big difference if I change versions?

[ornate elk]

whats your gpu?

[mortal compass]

NVIDIA Geforce RTX 3070 Ti, 8gb

32 gb ram, not sure that matters though

[ornate elk]

thats perfect to run sdxl models

[ornate elk]

should work good with auto1111

forge makes sense for 6gb or less vram

wouldnt use it with 8gb

it makes more problems that the little speed increase helps

[mortal compass]

Okay, I see. Maybe I'll just do it and follow your guide. Need to find a way to save the stuff I want to keep first

[ornate elk]

you dont need to delete forge for auto1111

you also can share models between them

[mortal compass]

That's more of a memory issue

[wet flame]

can you que up to batch processing?

[cursive cobalt]

What on earth is going on here?

Its ignoring both a negative, and a positive promt as well

[ornate elk]

Wrong vae

[cursive cobalt]

Wrong? Eh?

[ornate elk]

Sdxl and 1.5 models need a compatible vae

They can't use the same

[cursive cobalt]

OOh

Going the Lora way now, or trying at least, and the checkpoint way xD I be back if that doesn't work either

[cursive cobalt]

There a way to bookmark specific promts?

[mortal compass]

You need extensions for that. Or a really big txt document 😄

[cursive cobalt]

Right! Which extension, and compatible with Auto1111? 😄

[mortal compass]

I'm using this https://github.com/namkazt/sd-webui-prompt-history

it works with forge, so should work with a1111 afaik

[cursive cobalt]

https://tenor.com/view/thank-you-sticker-thanks-sticker-line-sticker-rabbit-sticker-bunny-sticker-gif-26476684

[mortal compass]

gozaimasu

[tawny flint]

I'm getting this error about facexlib. I am using python 3.10.11

Traceback (most recent call last):
File "D:\stable-diffusion\ComfyUI_windows_portable\ComfyUI\nodes.py", line 1879, in load_custom_node
module_spec.loader.exec_module(module)
File "<frozen importlib._bootstrap_external>", line 940, in exec_module
File "<frozen importlib._bootstrap>", line 241, in call_with_frames_removed
File "D:\stable-diffusion\ComfyUI_windows_portable\ComfyUI\custom_nodes\PuLID_ComfyUI_init.py", line 1, in <module>
from .pulid import NODE_CLASS_MAPPINGS, NODE_DISPLAY_NAME_MAPPINGS
File "D:\stable-diffusion\ComfyUI_windows_portable\ComfyUI\custom_nodes\PuLID_ComfyUI\pulid.py", line 10, in <module>
from facexlib.parsing import init_parsing_model
ModuleNotFoundError: No module named 'facexlib'

Cannot import D:\stable-diffusion\ComfyUI_windows_portable\ComfyUI\custom_nodes\PuLID_ComfyUI module for custom nodes: No module named 'facexlib'

i used pip to install facexlib

C:\Users\Eamon>D:\stable-diffusion\ComfyUI_windows_portable\python_embeded\python.exe -m pip install facexlib
Collecting facexlib
Using cached facexlib-0.3.0-py3-none-any.whl.metadata (4.6 kB)
Collecting filterpy (from facexlib)
Using cached filterpy-1.4.5.zip (177 kB)
Preparing metadata (setup.py) ... error
error: subprocess-exited-with-error

× python setup.py egg_info did not run successfully.
│ exit code: 1
╰─> [7 lines of output]
Traceback (most recent call last):
File "<string>", line 2, in <module>
File "<pip-setuptools-caller>", line 34, in <module>
File "C:\Users\Eamon\AppData\Local\Temp\pip-install-13q0izce\filterpy_456c11648240483cbf0da1cb5e214a09\setup.py", line 18, in <module>
version=filterpy.version,
^^^^^^^^^^^^^^^^^^^^
AttributeError: module 'filterpy' has no attribute 'version'
[end of output]

note: This error originates from a subprocess, and is likely not a problem with pip.
error: metadata-generation-failed

× Encountered error while generating package metadata.
╰─> See above for output.

note: This is an issue with the package mentioned above, not pip.
hint: See above for details.

[solemn atlas]

I don't know what to do

The majority is from SD

[lime dune]

#1237459938901491852

[ornate elk]

its talking about the C: drive
first open up a cmd and run
pip cache purge
Then get a tool like WizTree Free and scan your C drive for stuff thats taking up space

[viscid gate]

hi i cant get webui bat to open python

guys my PATH is set to 3.10

3.10.11

yet

whenever I open webui bat it says 3.11 could not be found

how do i change this

[cursive cobalt]

How does one purge the ram being used? Is a bit excessive

[viscid gate]

No python at

C:/user

what do i do

it keeps going to python11

[ornate elk]

Delete the venv folder and relaunch the webui-user.bat

[viscid gate]

ty

and python 3.10.11 is the correct version right

as u said in ur guide

[ornate elk]

Yep

[viscid gate]

@ornate elk

venv "C:\stable\stable-diffusion-webui-master\venv\Scripts\Python.exe"
fatal: not a git repository (or any of the parent directories): .git
fatal: not a git repository (or any of the parent directories): .git
Python 3.10.11 (tags/v3.10.11:7d4cc5a, Apr 5 2023, 00:38:17) [MSC v.1929 64 bit (AMD64)]
Version: 1.9.3

what dpoes this mean?

[ornate elk]

that means you didnt installed it correctly, meaning you didnt cloned it with Git
did you downloaded it as zip?

[viscid gate]

i did download it as zip

• Error loading script: main.py
  Traceback (most recent call last):
  File "C:\stable\stable-diffusion-webui-master\modules\scripts.py", line 508, in load_scripts
  script_module = script_loading.load_module(scriptfile.path)
  File "C:\stable\stable-diffusion-webui-master\modules\script_loading.py", line 13, in load_module
  module_spec.loader.exec_module(module)
  File "<frozen importlib._bootstrap_external>", line 883, in exec_module
  File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed
  File "C:\stable\stable-diffusion-webui-master\extensions\openpose-editor\scripts\main.py", line 14, in <module>
  from basicsr.utils.download_util import load_file_from_url
  ModuleNotFoundError: No module named 'basicsr'

[ornate elk]

yea thats not the correct way, my guide has the correct steps to install it

[viscid gate]

i tried cloning and ti says

$ git clone github.com/AUTOMATIC1111/stable-diffusion-webui
fatal: repository 'github.com/AUTOMATIC1111/stable-diffusion-webui' does not exist

[ornate elk]

the link is incorrect its missing https://

you need:
git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git

[viscid gate]

oh

[ornate elk]

openpose editor is broken since last year as it didnt got an update, the extension can be fixed by installing basicsr manually or can be deletet because controlnet has an openpose editor too that works

[viscid gate]

same error after i cloned it into new folder

[ornate elk]

can you show it?

[viscid gate]

nvm i was opening wrong folder

[viscid gate]

@ornate elk

i cant install

whenever i install i just get errors

this is on a clean inmstall in a new folder

[ornate elk]

how much RAM do you have?

and how much free Space on the C: drive?

[viscid gate]

16gb ram and around 20gb free on c drive

[ornate elk]

okay make sure to close every other programm on the PC, the installation needs a lot of RAM

[viscid gate]

k

[ornate elk]

also maybe restart the PC before trying again

if it still crashes you need to increase your windows pagefile (or check if you miss it)

[viscid gate]

File "C:\Users\Zain Raza\AppData\Local\Programs\Python\Python310\lib\threading.py", line 1016, in _bootstrap_inner
self.run()
File "C:\stable\stable-diffusion-webui\venv\lib\site-packages\anyio_backends_asyncio.py", line 807, in run
result = context.run(func, *args)
File "C:\stable\stable-diffusion-webui\venv\lib\site-packages\gradio\utils.py", line 707, in wrapper
response = f(*args, **kwargs)
File "C:\stable\stable-diffusion-webui\modules\ui.py", line 1154, in <lambda>
update_image_cfg_scale_visibility = lambda: gr.update(visible=shared.sd_model and shared.sd_model.cond_stage_key == "edit")
File "C:\stable\stable-diffusion-webui\modules\shared_items.py", line 175, in sd_model
return modules.sd_models.model_data.get_sd_model()
File "C:\stable\stable-diffusion-webui\modules\sd_models.py", line 620, in get_sd_model
load_model()
File "C:\stable\stable-diffusion-webui\modules\sd_models.py", line 748, in load_model
load_model_weights(sd_model, checkpoint_info, state_dict, timer)
File "C:\stable\stable-diffusion-webui\modules\sd_models.py", line 393, in load_model_weights
model.load_state_dict(state_dict, strict=False)

out of memory

torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 20.00 MiB. GPU 0 has a total capacty of 8.00 GiB of which 3.19 GiB is free. Of the allocated memory 3.68 GiB is allocated by PyTorch, and 94.77 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting max_split_size_mb to avoid fragmentation. See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

[cursive cobalt]

Mmmm Out of memory. THat is what i live for.
Not even regular ram (32gb), can keep up. (12 gb VRAM).

[viscid gate]

im literally just opening

web ui

and its out of memory

[cursive cobalt]

@ornate elk Has u told him to do the low/medium vram tweak in the bat?

[ornate elk]

whats your gpu?

[viscid gate]

2060s

[ornate elk]

and what model is in stable-diffusion-webui/models/stable-diffusion

[ornate elk]

okay and whats in your webui-user.bat?

[viscid gate]

base model

[viscid gate]

i put ur settings

--medvram

[cursive cobalt]

That is dangerous. Never know if they are compatible

[ornate elk]

you need --xformers --medvram-sdxl --no-half-vae

[viscid gate]

--xformers

-no-half-vae

i hvae it

oh i didnt add sdxl

[ornate elk]

and you get the error when pressing generate ? or just by opening the webui in browser

make sure to always launch the webui-user.bat and not the webui.bat

[viscid gate]

generate

[ornate elk]

oh what resolution?

and steps etc

[viscid gate]

512x512

20 steps

base model

v1.5

[ornate elk]

hmm and nothing activated?

[viscid gate]

nope

fresh install

[ornate elk]

thats really strange

[viscid gate]

no scripts

i may have found the culprit

adobe creative cloud

[ornate elk]

anything running in the bckground? like epic game launcher or wallpaper engine?
or hundreds of browser tabs open?

[ornate elk]

ohhhh

this thing is an resource hungry tool

[viscid gate]

it was using like 7gb ram

idk why

i uninstalled it

[ornate elk]

ohhh wow

xD

[cursive cobalt]

Can you sit the default resolution, the webui will always open up with? This is more in case of a crash, and i forget to re-do that specific setting

set the*

[ornate elk]

yep if you edit the value in ui-config.json

[cursive cobalt]

Excellent

[frozen parrot]

hey if anyone could help, I seem to be getting some very bad images, but with no errors.

[timid notch]

haha,i find something funny~

i got this error early and dont know how to fix it.

today i restart my computer

everything is ok~~so the bast way to fix a problem is restart?

[cursive cobalt]

Could try to specifiy the amount of each gender the scene has? 1 boy, 1 girl, 2 boys, 2 girls, etc, and see if that fixes? Am also really new to this, but i use those promts frequently.

[frozen parrot]

I've tried that as well, it seems like the image is just being fried, as far as I can tell this isn't a prompt issue

[ornate elk]

Denois to high, set it to 0.5 and hires steps to 10

[final latch]

Hey yall, I've been trying to diagnose and fix a really persistent and damaging issue, that's got multiple levels of weird behaviour, but always has the same outcome where the final generation is just completely toasted - or turns into some sort of horrorscape. I've tried everything from fiddling with Nvidia drivers, changing args, updating python, reinstalling auto1111, using comfyui, even waiting a month for new drivers and updates - nothing seems to stop my pc from just frying the image. At this point I'm contemplating if my graphics card is just busted, and I simply don't notice it while gaming.

Sometimes I'll get a barely usable image (even though the generation completely ignores half the prompt), and then for each generation it just declines in quality - never reaching the actual potential of the model. - There should be a paladin with a bucket on his head in this image, half of the prompt is dedicated to his description, but all it spits out is vague scenery.

[thorny iron]

Hello

I reinstalled my OS and cannot remember what I'm doing

Stable Diffusion is 'installed' but I'm missing Python

What version do I need

I remember latest is not a good idea.

[cursive cobalt]

... Why isn't the latest a good idea?

Oh right, incompatibility. 2 sec, i installed this stuff like 2 days ago

3.10.6, i believe was the version u wanted

[thorny iron]

That's the one thank you.

[remote ivy]

I am having an issue with automatic1111 where it is using close to 100% of my system ram all the time. Does anyone else have this issue?

[thorny iron]

How much RAM you have

@remote ivy

[cursive cobalt]

Yeah... Mine will frequent end up capping out on my ram (32gb), as well.

[remote ivy]

Is there anyway to mitigate this? I am coming to automatic1111 from forge where I was not experiencing this issue.

My reason for switching is issues I was running into with forge's built in controlnet version

[thorny iron]

3.10.11 is the latest you can have it seems just FYI

[cursive cobalt]

Fairsies

[thorny iron]

.. Well this sucks.

[cursive cobalt]

At least you get an error code

[thorny iron]

I'm assuming it's because I'm using an old install and my prior user account had a specific token associated with it or something

I'll just make a new one, oh well

ayyy

[remote ivy]

32 gb of ram

[thorny iron]

Me too

I think I've been maxed a couple times too

I just reboot the program.

I think it happens when you interrupt a job or something gets stuck

[cursive cobalt]

Why is it spamming this, despite what seems to be it reading the tags? (Forest and landscape was added AFTER that grid)

[thorny iron]

Ayy it works.

See you guys in 4-6 months when I format again

[scarlet crater]

Anyone else having trouble disabling soft inpainting via API?

[cursive cobalt]

Kek. Electri Boogaloo

[spice gazelle]

been having a weird problem since the last git pull update from yesterday, after everything loads correctly from the webui.bat and it auto opens web ui it will sit there frozen/loading with timers above the checkpoint/clip and anything else pinned to the top for 500s then error connection pops up like 6 times then its useable like normal. happens every time if i close the window and reopen a new one from the local url. have to wait 500s if i reload the ui to change something have to wait the 500s. Was fine before the update so idk what happened

[woven apex]

ive installed reactor extension but it doesnt swap faces what problem can be?

[ornate elk]

Hey, can you show your txt2img settings? Also what's your GPU?
To start fixing stuff open up a cmd and type
SFC /scannow
That will fix corrupt system files

[ornate elk]

Missing insight face

[ornate elk]

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

[woven apex]

Where can I download the missing file?

[remote ivy]

This is just what it will normally use for me regardless of if it interrupts or not.

[ornate elk]

You have to follow the instructions by Reactor:
https://github.com/Gourieff/sd-webui-reactor?tab=readme-ov-file#installation

[final latch]

Sure thing!
I run everything on a gigabyte 3090ti
Parameters for the weird minimalist cubic failed image:

score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, (close-up) of a large radiant human paladin, wearing (huge plated armor, bucket helmet, bucket on head, large shoulder pads, blue cloak on back), holding a stick in his hand, striking a pose, mountain peak, sunny day, clear skies, castle in background, concept art, realistic, fog, volumetric lighting, depth of field, macro lens, <lora:Concept Art Twilight Style SDXL_LoRA_Pony Diffusion V6 XL:0.8>, lora:xl_more_art-full_v1:0.6
Negative prompt: face, food
Steps: 25, Sampler: Euler a, Schedule type: Automatic, CFG scale: 7, Seed: 3675776181, Size: 1344x896, Model hash: 67ab2fd8ec, Model: ponyDiffusionV6XL_v6StartWithThisOne, VAE hash: 235745af8d, VAE: sdxl_vae.safetensors, Clip skip: 2, Lora hashes: "Concept Art Twilight Style SDXL_LoRA_Pony Diffusion V6 XL: e5fe96cd307b, xl_more_art-full_v1: fe3b4816be83", Version: v1.9.3

I'll try the SFC /scannow

[spice gazelle]

4070 super and no commandlines inside it execpt for git pull and pause at the top

[ornate elk]

Then you should add --xformers --no-half-vae to the Commandline_args=
Also make sure to always launch the webui-user.bat

And not the webui.bat

And if you still get an error check what model it tries to load

[ornate elk]

Ah pony model. These models can break easily by using a wrong word, not compatible lora or a wrong setting in auto1111

[final latch]

Yeah, I've noticed the model breaking quite often - but that usually appears like a completely different kind of artefacting. This is just straight up every single model - both sd and sdxl getting crunched.
It also just happened quite suddenly, and has meant a complete non-progress of any stable diffusion on y pc

[ornate elk]

And your using a fresh Auto1111 install?

[cursive cobalt]

Is there a plugin, that allows you to re-name Lora's, in the Lora list, without having to go into the Lora folder to locate it?

[ornate elk]

Maybe try a normal sdxl model without using any lora or embedding

[final latch]

Yep

[final latch]

Tried completely fresh install, re-downloaded all models and loras, tried without loras... all sorts

[spice gazelle]

I use to have those in the command lines but tried uninstalling and reinstalling to see if it would fix it so currently didn't have them in the commands problem happens with either, but if i leave the url open after its already loaded i don't have to wait

[thorny iron]

Sometimes I get this

[final latch]

Tried to run the SFC /scannow - it found some corrupted files and repaired them but results are still the same

[thorny iron]

What can I do to either fix it, or abandon my current generation and moveo n

[cursive cobalt]

Lemme guess, it gets stuck on 4s?

[thorny iron]

Well

[cursive cobalt]

And slowly adds a second eveyr ~5 sec?

[thorny iron]

... :'c

Cries

[cursive cobalt]

Was i guessing correct?

[thorny iron]

Cries more

... Oh it finished it finally, wew. Gangsign Peace.

[cursive cobalt]

CRYING IS NOT AN ASWER!

[thorny iron]

It's an admission of the painful truth and reality

[cursive cobalt]

Kek

Mine does the same

I let it sit. But because my gpu is a bit old, if it surpasses an added 60 sec to the timer, I assume its frozen

And reboot the webui.

I switched mine to run in Firefox... Well, I've made it my home page, on firefox

[ornate elk]

Also check out a different browser. Someone had problems in chrome since an update

[ornate elk]

What happens when you use an 1.5 model?

[spice gazelle]

i use brave which runs on chromium tested out edge real quick seems to be the problem is chrome based browser

[ornate elk]

Did it worked in Firefox?

[final latch]

I can generate a 512x512 image mostly fine (bear in mind i don't use 1.5 much since switching to sdxl, so i don't know whether the quality has tanked) but slowly for each image generation it becomes more unstable (like the sdxl images i showed) and if i do a batch generation of 8 images at once (something that my pc used to be perfectly fine with, never had any problems) i can literally see how it's completing less and less images for each generation, using the same parameters every time until it spits out vram errors

[thorny iron]

I'm trying to replicate a picture from CIVITAI but some reason my output is a lot different

[cursive cobalt]

You could try to pull the meta data from png info?

[thorny iron]

Is it because I don't know what these are and I'm putting them in the wrong location?

I think I'm adding them all as LORAs

I'm using Automatic1111 by the way

[cursive cobalt]

Checkpoint is the top left spot, Lora are added as lora ofc. Embeds go in the embed folder, X:\A1111\stable-diffusion-webui\embeddings

Lycoris... However, I'm not sure what, and how they work. CS1o would know

[thorny iron]

Idiot

[cursive cobalt]

Scuse me?

[thorny iron]

Of course I put them in the LORA folder olol

I'll put them in the other one and try again

Me not you

[cursive cobalt]

Ah, i c 🙂

[thorny iron]

CS1o is awesome, he's super helpful

[cursive cobalt]

He is!

https://tenor.com/view/cute-bunny-gif-26165471

[thorny iron]

PLEASE SEND ME ALL OF THOSE

Alright, Embeddings attempt 2 - Electric Boogaloo.

[cursive cobalt]

Lemme know when you find out what lycoris is, and how they work

[spice gazelle]

so upon further testing it looks like it wasn't any of the browsers chrome, firefox, edge all work its an extenstion i have in brave my main one and its the Malwarebytes browser guard extension that after the newest update is causing it to load in slow taking the 500s timeout connection before being useable. disabling it allows it to boot right up disabling causes the issue now why its causing an issue and why after the newst git pull update i have no clue lol

[thorny iron]

Well

I got these things like Realistic Vision, BetterFreckle etc that was spat out for prompts

[ornate elk]

Yes the webui has to be whitelised in any adblocker browser extension

[thorny iron]

Problem is, it seems I need to add them myself from the LORA tab.. but I don't know their weighting.

[cursive cobalt]

Weight:2000

Send it

[thorny iron]

LOL ;d

[spice gazelle]

is that a new change? because i've never had the issue before

[ornate elk]

Really strange

[thorny iron]

Neither and I use uBlock and stuff.

[cursive cobalt]

Imo, just get a seperate browser, and change that browsers home page, to the webui 😄

Allows a clean client, without any interference from anything

[thorny iron]

Goddammit it keeps going to tits.

[cursive cobalt]

Generating tits doesn't sound too bad... Unless its a male character

[thorny iron]

I wish the layout of information spat out from CIVITAI was easier to understand.

I mean tits as in.. it's bad. ';d

What drives me up the wall is so many LORAs and so on have confusing names, and don't show up the same way in the client.

[ornate elk]

It depends on the browser extension

[cursive cobalt]

Thats why i spent like 2 hrs yesterday, making preview thumbnails, and re-naming all of my LORA's

[thorny iron]

Gah, idiot again..

I should have started with 0.4

I thought 0.8 was enough

Maybe even less

[cursive cobalt]

Btw, @ornate elk Happen to know of an extension that allows me to re-name Lora's IN the webui?

[thorny iron]

Wait, is there a standard / default weighting when CIVITAI uses it?

[cursive cobalt]

Doesn't png info tab show all of this info, if you're trying to replicate an image?

[thorny iron]

Yes that'd be nice.. also how to make the LORA's either have pictures or smaller icons. I feel like I'm playing YU-GI-OH

Sadly it just says what's used, but not the weighting

Wait

..... Ooooh..

[cursive cobalt]

It should if you send it to txt2img

[thorny iron]

... I wonder if I can indeed just download it and shove it in there.

Gangsign

[cursive cobalt]

As long as its a png file, it should have all of the meta data it can draw information from

[thorny iron]

HELL-E-YEAH

[ornate elk]

There is a setting in the webui which makes it display the original file names, then you can rename it I guess

[thorny iron]

Wait nevermind, it still failed lol

I mean the prompt reverse worked, but the LORA info or SOMETHING is still missing..

[cursive cobalt]

You see, the issue is i May have named a few things, the same, with minor variations, such as Lora1, and Lora1,2

[thorny iron]

Is it possible some CIVITAI stuff is omitted from users or something unless specifically published?

[cursive cobalt]

Or Lora.1, lora1,

[thorny iron]

Wait

[cursive cobalt]

And i wanna clean up my lora folder, cause i have a few... invalid ones

[thorny iron]

What's 'Clip Skip'

I need to do the same with my models :f

Dear god.

[cursive cobalt]

Settings > Stable Diffusion > Clip Skip

Trying to find an easier way to do it

[cursive cobalt]

https://github.com/AUTOMATIC1111/stable-diffusion-webui/discussions/3004 here u go :3

[thorny iron]

What does it do? :f

[cursive cobalt]

No idea, but i see a crap ton of images from civit, using clip skip, so figured having it easily available would be beneficial :;D

[thorny iron]

Well god damn it this whole time this might've been my issue.

[cursive cobalt]

My python. AND taskmanager both crashed 😐

[thorny iron]

LOL

[cursive cobalt]

So Python crashed so bad, i can no longer close the cmd window

[thorny iron]

Usually what I do is I just press X, anywhere on the window, then click 'cancel' on the popup window

[cursive cobalt]

Yeah, the window that asks if i want to end the process, doesn't even work xD

I mean it pops up, but doesn't work

[thorny iron]

Yeah clicking cancel for some reason is what makes it proceed

At least for me. Like don't press 'OK'

[cursive cobalt]

Nop. Cancel just closes the box xD

[thorny iron]

It's like some weird reverse psychology thing

Weird.

I must have some tweak in my PC that set a registry entry to have like 0 wait time on hanging programs or something

Because the moment I press cancel it just kills it

[cursive cobalt]

Right. Nothing works, brb rebooting pc i guess

Or not... Reboot doesn't work either

Hard rebooting it then

[solemn atlas]

what does that mean

[thorny iron]

Humm

Are you still here

Try to do a safe one.. :<

:C

Yep

It's without a doubt the fact I don't know the correct LORA weightings.

How annoying. CIVITAI doesn't say what they are.

CS1oooo!!!!!

[cursive cobalt]

I have returned. FOr some reason, my gpu refuses to display over displayport, after a hard crash

So gotta unplug it, wait for it to boot up via the hdmi, then reboot and plug DP back in

[thorny iron]

:c

[cursive cobalt]

Its great

./s

[thorny iron]

Unplug what

[cursive cobalt]

over Displayport, after a hard crash, so i gotta unplug it 😉

[thorny iron]

Apparently you should unplug your screen

And let it sit 60 seconds :<

[cursive cobalt]

No no, that won't work

[thorny iron]

As in, the power part.

:C

[cursive cobalt]

Just gotta remove the cable from the gpu

Ez

(Is sitting on the desk /w me)

[thorny iron]

Well I give up. :C

.. Stupid LORA weightings

You suck ;U

I'll try again later.. maybe we can learn how they work better sometime together. :x

I just figured out how to fix this and my results changed crazy.

[ornate elk]

Incompatible stuff used together

[mortal compass]

Btw @ornate elk my issues got solved by just deleting the ui-config file and move my wildcards before civitai+ updates them and adds html's and pictures. (???) lol

[ornate elk]

ahh okay

[mortal compass]

Will git checkout dev fuck up my forge you think 🤔

[ornate elk]

it probably does

its dev

[mortal compass]

I HAVE LEARNED! To actually ask someone first before I do something that I have a bad feeling about

Feelsgoodman

Man I wanted to try that sweet euler a karras

[solemn atlas]

is this controlnet?

[ornate elk]

No

Its called sd-webui-controlnet

[shell goblet]

Hi! I'm in the "--skip-torch-cuda-test" bucket of people, is there another way to force it to use my GPU? As far as I understand this makes it use my CPU, and it's slow. I'm waiting 5-ish minutes for an image.

[cursive cobalt]

:CriesIn1080TI: @ 1-2 min 😦

[ornate elk]

Hey, what's your GPU?

[pallid bramble]

Hey guys, what's the deal on using the API for adult content? Is that a content policy issue?

[shell goblet]

AMD Radeon RX 7700 XT

[ornate elk]

Then you need to follow my AMD Zluda webui install guide.
Its in the pinned messages of this channel

With that you can use Stable-diffusion on your GPU

[shell goblet]

You're a good person, I appreciate you!

[ornate elk]

No problem 🙂
Feel free to ask if you have any questions

[cursive cobalt]

@ornate elk Friend! I think i made a boo boo.

I installed https://github.com/lllyasviel/ControlNet-v1-1-nightly first, then after reading (After importing it), that its NOT for A1111. I then, installed the correct version (https://github.com/Mikubill/sd-webui-controlnet), and restarted the UI... Now i can't boot it back up

[dusk lantern]

Does anyone use Replacer addon with SDNext?

[ornate elk]

that pydantic error appears when using a not updated controlnet version.
Can be resolved by doing this:
Go into the stable-diffusion-webui folder. Click in the explorer bar (not search bar)
Type in cmd and hit enter.

Run the three commands one by one:

venv\Scripts\activate pip install albumentations==1.4.3 pip install pydantic==1.10.15

Then relaunch the webui-user.bat

[cursive cobalt]

Must've made a typo, this time it installed

Is still not liking it @ornate elk 😦

Running the cmd's again, yields this

[cursive cobalt]

Beep boop @ornate elk (<3)

(Sorry for the excessive pings 😦 )

[lunar glade]

Hi, I'm getting this error with Kohya. What can I do?
File "E:\AI\Kohya\kohya_ss\sd-scripts\sdxl_train_network.py", line 185, in <module>
trainer.train(args)
File "E:\AI\Kohya\kohya_ss\sd-scripts\train_network.py", line 272, in train
train_dataset_group.cache_latents(vae, args.vae_batch_size, args.cache_latents_to_disk, accelerator.is_main_process)
File "E:\AI\Kohya\kohya_ss\sd-scripts\library\train_util.py", line 2080, in cache_latents
dataset.cache_latents(vae, vae_batch_size, cache_to_disk, is_main_process)
File "E:\AI\Kohya\kohya_ss\sd-scripts\library\train_util.py", line 1023, in cache_latents
cache_batch_latents(vae, cache_to_disk, batch, subset.flip_aug, subset.random_crop)
File "E:\AI\Kohya\kohya_ss\sd-scripts\library\train_util.py", line 2428, in cache_batch_latents
raise RuntimeError(f"NaN detected in latents: {info.absolute_path}")
RuntimeError: NaN detected in latents: E:\AI\Regularizacion\Maru IA\1_ohwx\ohwx (1).png
Traceback (most recent call last):
File "C:\Python\Python310\lib\runpy.py", line 196, in _run_module_as_main
return _run_code(code, main_globals, None,
File "C:\Python\Python310\lib\runpy.py", line 86, in run_code
exec(code, run_globals)
File "E:\AI\Kohya\kohya_ss\venv\Scripts\accelerate.EXE_main.py", line 7, in <module>
File "E:\AI\Kohya\kohya_ss\venv\lib\site-packages\accelerate\commands\accelerate_cli.py", line 47, in main
args.func(args)

File "E:\AI\Kohya\kohya_ss\venv\lib\site-packages\accelerate\commands\launch.py", line 1017, in launch_command
simple_launcher(args)
File "E:\AI\Kohya\kohya_ss\venv\lib\site-packages\accelerate\commands\launch.py", line 637, in simple_launcher
raise subprocess.CalledProcessError(returncode=process.returncode, cmd=cmd)
subprocess.CalledProcessError: Command '['E:\AI\Kohya\kohya_ss\venv\Scripts\python.exe', 'E:/AI/Kohya/kohya_ss/sd-scripts/sdxl_train_network.py', '--config_file', 'E:/AI/New Training Data\model/config_lora-20240519-191339.toml']' returned non-zero exit status 1.
19:15:40-249264 INFO Training has ended.

[wary maple]

Trying to crop my img but this happen 😦

[cursive cobalt]

#🤝｜tech-support message Fixed it by deleting the Venv folder, and launching the Webui.bat file to reinstall the venv folder

[wary maple]

I just create an embedding for training a new model but there's nothing in the embedding drop list, ive restarted and refreshed it. There's jsut nothing 😦

[wary maple]

can anyone help with this🥺

[willow kettle]

Alright, another edge case with my AMD gpu. Is there a way to run rembg (remove background) in A1111 on cpu instead of it trying to use gpu? It keeps trying to use cudnn...

[wet flame]

foocus running slower on my new gpu 16 gb than my 4 gm one lol

[kindred wolf]

Is it possible to get support for Stability Platform API from somebody from the team here? Been trying to reach out over email and helpdesk with no success. We got some billing issues which prevent us from giving money to Stability 🙂

[jovial ermine]

Hi everyone, I am trying to install SD and I am getting this error

RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check

What is the best way to solve? I am quite new to this so plrease excuse my ignorance. Many thanks in advance...!

[untold flame]

@woeful pendant Are you here by chance?

I have been having a VERY windows specific error for Months and you are the only one I remember in the past being able to help me fix it. It has been driving me absolutely insane because it makes literally negative sense. If you could spare a little time, I would appreciate it immensely

Its not specific to SD, just installed programs as a whole (been affecting LLM's the most)

[woeful pendant]

yes

[untold flame]

Would you be willing to take a look? I have exhausted so many options and it makes no logical sense at all. I would be immensely grateful 😅

[woeful pendant]

what is happening or it doing it shouldn't?

[untold flame]

its such a simple/stupid error

this has been haunting me for MONTHS

I can't build anything at all on my system

It tells me it can't find it, but then when I ask it to print my CUDA_HOME var, it prints it

This has literally been ruining everything I do with AI for well over half a year, and I can't find any information on how to fix it or what would even cause it

I have reinstalled python, cuda, my GPU drivers, VSC, C++, everything so many different times, and its never fixed or even changed it. Its just never worked on this PC

Cuda 11.7/11.8/12.0/12.1/12.2 all do it

Same with python 38/39/310/311

[woeful pendant]

I don't use that but in a cli type echo %CUDA_HOME%

[untold flame]

wait...

[woeful pendant]

remove the type, lol

[untold flame]

type?

[woeful pendant]

as in typing just type echo %CUDA_HOME%

[untold flame]

oops lmao

[woeful pendant]

there it is

[untold flame]

I even tried manually setting it for the operation in windows using the set function, but it STILL tells me its undefined

[woeful pendant]

have you installed the cuda toolkit?

[untold flame]

over 10 different times

clean installs

I can re-install again

[fair oxide]

Hello, anyone knew how to install stable diffusion on computer?

[woeful pendant]

You use conda?

[untold flame]

I do not use conda, but it does seem to work inside conda for whatever reason. Though I have some programs that specifically say they are not compatible with running in conda

auto install scripts work fine

but if I run the individual commands manually to install in my own venv, it never works

[woeful pendant]

I forget how censorious this discord is but I did find this

As I think other people may end up here from an unrelated search: conda simply provides the necessary - and in most cases minimal - CUDA shared libraries for your packages (i.e. a bunch of .so files). I don't think it also provides nvcc so you probably shouldn't be relying on it for other installations. (I ran find and it didn't show up).

If you need to install packages with separate CUDA versions, you can install separate versions without any issues. The downside is you'll need to set CUDA_HOME every time.

The error in this issue is from torch. So you can do:

conda install pytorch torchvision cudatoolkit=10.1 -c pytorch

and it should load correctly.

Just revise to current versions.

that last line is me. conda and torch can be problematic

[untold flame]

unfortunately I need NVCC for a lot of these

[woeful pendant]

conda has its own evironment variables

[untold flame]

I just don't understand why windows can't seem to find a file it points directly at

[woeful pendant]

it isn't windows at fault here

[tiny garden]

It seems I have a problem with my A1111:
I switched my base model from SD1.5 to SDXL and wanted to select the Fantasy Warriors LoRA - only to find out it isn't displayed there despite it being located in the matching folder looking in the Windows Explorer. Is that a display bug that can be solved with a simple re-start of A1111 or is that a deeper problem?

[woeful pendant]

it is Conda it can't find it

[untold flame]

What do you think is at fault in this case? I am at a loss

[untold flame]

oh, this is not using conda

[woeful pendant]

yes it is

[untold flame]

I have other programs that do use conda, but this doesn't

[woeful pendant]

ahhh, llama

type nvcc --version

[woeful pendant]

you keep typing type

[untold flame]

DAMN IT

type again lmao

sorry, sorry 😅

[woeful pendant]

I never went past 12.1 as it had issues

[untold flame]

When I try to build files, they all error our in conda, venv, and base PC saying they can't find cuda and that I do not have cuda installed

And some packages that I download say they can't find the DLL while looking in the exact folder with the DLL. If I disable the check, they work fine

[woeful pendant]

see this? there is your error

Python, but I need to see that file

[untold flame]

I have had this on 11.7/11.8/12.0/12.1/12.2

[untold flame]

Alright, I can send it

[woeful pendant]

k