🤝｜tech-support

[urban girder]

now the next obstacle, making accurate and good anime art.

[ornate elk]

I'm off for today(night) but feel free to ask in #📝｜prompting-help or checkout civitai.com for anime models

[urban girder]

you've done so much to help me. Thank you. No need to go further this is a huge landmark for me.

[ornate elk]

NP have fun!

[steady thistle]

Have a great night!

[ruby yoke]

try putting this phrase "Yoshiyuki Tomino anime " at the very start of your prompt (remove the " " marks)

[urban girder]

do I put any commas?

[ruby yoke]

nope

just put those three words, then put the rest of your prompt. works extremely well with flux - should work fine with SDXL

[urban girder]

ok does this prompt look good?

orcs with green skin and attractive faces wearing brown ww1 trench coats medieval chest armor and shoulder armor marching through no man's land Yoshiyuki Tomino anime

[ruby yoke]

Yoshiyuki Tomino anime orcs with green skin and attractive faces wearing brown ww1 trench coats medieval chest armor and shoulder armor marching through no man's land

start of the prompt, not end of the prompt - stuff at the start has more weight than stuff at the end

you might also try this prompt: Yoshiyuki Tomino anime orcs with green skin and attractive faces. They are wearing brown ww1 trench coats over top of medieval chest armor and shoulder armor. They are marching through no man's land

(remember, just because you can describe something that is underneath of and covered up by something else, it doesn't mean you - or the AI - can draw it

[azure magnet]

An help to fix it?

[ruby yoke]

it's not finding insightFace - is it installed?

[azure magnet]

are you skilled in trouble solutions, even if i think yes, how can i check?

[oak jay]

Hi all, I installed this https://github.com/harukei-tech/sd-webui-extended-style-saver, and I keep getting an error, any ideas how to fix it please?

[ruby yoke]

are you using auto1111, comfyUI, webUI, somethine else?

[ruby yoke]

what error are you getting?

[azure magnet]

comfyui

[oak jay]

Traceback (most recent call last):
File "C:\Users\Philip\Forge\webui\venv\lib\site-packages\gradio\queueing.py", line 536, in process_events
response = await route_utils.call_process_api(
File "C:\Users\Philip\Forge\webui\venv\lib\site-packages\gradio\route_utils.py", line 285, in call_process_api
output = await app.get_blocks().process_api(
File "C:\Users\Philip\Forge\webui\venv\lib\site-packages\gradio\blocks.py", line 1919, in process_api
inputs = await self.preprocess_data(
File "C:\Users\Philip\Forge\webui\venv\lib\site-packages\gradio\blocks.py", line 1647, in preprocess_data
inputs_cached = block.data_model(**inputs_cached) # type: ignore
TypeError: gradio.components.label.LabelData() argument after ** must be a mapping, not str

[ruby yoke]

load comfy, go into manager, go into the node manager, search on insightface and see if it's installed

[ruby yoke]

can you post the entire error log please

[oak jay]

That is it

[ruby yoke]

that's all that's on your console screen?

[azure magnet]

there isn't a any node called insightface

[ruby yoke]

here's the github for insightface https://github.com/deepinsight/insightface

[ruby yoke]

and you didn't just isntall something new?

[oak jay]

Just that extention that I mentioned

Only the extension is faulting

Everything else is working

[azure magnet]

Are you sure that i have to install it?

[ruby yoke]

if you're sure it installed correctly, i'd consider a possible memory corruption issue and try a reboot

[ruby yoke]

the error says that the package is trying to find it. i don't know if you have to install it or not, but if it was me, i would

[oak jay]

I installed from the extensions tab at first, when that didn't work I deleted it the reinstalled it by pasting the git repository URL into the 'install from URL' field.

And I've done a reboot

[ruby yoke]

i tend to avoid that and follow the install instructions on the github page - usually that starts with a git clone and then other stuff

[oak jay]

@ruby yoke The instructions said to use 'install from URL' which is why I tried it that way the second time. Do you know of any other way of saving good prompts?

[snow breach]

Does anyone happen to know if kohya_ss can pickup training from the last saved checkpoint? I stopped training between epoch 3 and 4. Saved my settings. Restarted my computer. Loaded settings and clicked train. It started at epoch 1 step 0. Wondering what I have done wrong?

[snow breach]

Oh. How embarassing.

[median geode]

fail again. what does it mean?

should i add a space before --medvram? does it matter?

[ruby yoke]

yeah. those are command line arguments and they need spaces between each of them

[gloomy bolt]

I jsut wanted to check if it's normal for the next generation on Ponyxl (auto1111-> ForgeUI) to take an insane amount of ram after switching checkpoints

Everytime I swap a model, and wait for it to finish, then generate an image, it would take like 6 minutes to finish (usually 10-20 seconds), and i'm assuming it's using a lot of ram because my whole PC lags

but if i interrupt it and generate again the issue disappears

[craggy epoch]

I always assume the first image will take longer than usual (but I was using an ARC 770 until recently, so it's definitely normal there)

[gloomy bolt]

i'm on a 4070, and i hardly change models so i couldn't really tell if it's normal or not

[craggy epoch]

I haven't paid attention since moving to a 4070, I'll have to fire it up and check... Windows or Linux?

[gloomy bolt]

windows

[craggy epoch]

I might be swapping back to Win.... can check it on mine after I do so, or if someone else has a similar setup they can check in the meantime

[gloomy bolt]

np take ur time, it's not anything urgent, just want to make sure somethings not wrong w my pc or sd

[craggy epoch]

yeah, I would say anecdotally, the first image has always taken longer and with fewer it/s than others in a batch

[ruby yoke]

it has to load the new model into memory when you swap it out, so most of that time is just loading the model. once it's loaded, the gen time is just the image generation

[gloomy bolt]

hm i see

so just give it a min

[ruby yoke]

yup. just let it run

[median geode]

what should i do next? is webui already downloaded?

[ornate elk]

What's your GPU?

[median geode]

Intel(R) UHD Graphics 630 and NVIDIA GeForce GTX 1650

[ornate elk]

Make sure your nvidia driver is updated
Then relaunch again

[fair oxide]

witch cogvideox model is good for a 4060ti 16vram? for comfy ui

[median geode]

after updating n driver, a new error just come

i've almost came up to all kinds of errors in downloading webui

[ornate elk]

Https error is a connection error

Try use a VPN or try not to use a vpn

And try change your DNS

[wild finch]

this error is coming

i have correct py version

how to solve

[ornate elk]

Wrong Python Version.

Uninstall 3.12

Then install 3.10.11 64bit

Then delete the venv folder and relaunch the webui-user.bat

[wild finch]

i dont know where in venv folder

and i have 3.10.11 64 bit

[ornate elk]

Delete the venv folder

Its in stable-diffusion-webui

[wild finch]

thanks

@ornate elk u r goated

[ornate elk]

^^ thx and np

[wild finch]

whre do i find the webui after download i am searching some amt of time

[ornate elk]

Its in the folder where you git cloned it into

[wild finch]

what to do

[wild finch]

k

[ornate elk]

What's your GPU?

[wild finch]

AMD Radeon HD 7670M

sry for late

@ornate elk

[cosmic furnace]

which one of these zluda versions are best & recommended ?

Automatic1111 with ZLUDA
Automatic1111 with DirectML
Comfyui with ZLUDA
Comfyui with DirectML
SwarmUI with ZLUDA
SwarmUI with DirectML
Fooocus with ZLUDA
Fooocus with DirectML
SD.Next with ZLUDA (soon)
AMD Shark Studio (Currently Broken)
Forge-Webui with ZLUDA

ive heard comfy faster than automatic and best creators prefer it ? hm

[median geode]

what does it mean this time tired of it

[ornate elk]

"Unexpected disconnect" means unstable connection

[median geode]

🥲 error all the time

[median geode]

what does it mean 'Couldn't determine Stable Diffusion's hash'

[ornate elk]

strange error

but its from git

it cant check if the repo is correct

[median geode]

what should i do now🥲 maybe it's because i updated N driver today?

[ornate elk]

no its still related to the connection to github

did you changed your DNS server?

that could help

or try to reinstall Git

[median geode]

whats that

maybe i should reinstall git

[median geode]

im downloading git and which one should i choose

[ornate elk]

you can leave it as it is

[median geode]

git reinstalled, still the same

i think some document might disappear after i unintall the old git

[solemn atlas]

Hello guys, as we talked yesterday. I try to move everything to my (C:) drive which is my SSD.

Now I get this error code:

error code 0x8007080

[median geode]

show the details

[ornate elk]

You can't move that as it has 0 bytes

[median geode]

help！help! what should i do now

[mystic lava]

@ornate elk Yo!

You helped me a ton a whilke ago

is there a faaster version of flux yet for amd cards?

[ornate elk]

hey, what was your gpu again?

[mystic lava]

Rx 7900 xt

:D

[ornate elk]

i tried the gguf flux models recently

they need less vram

[mystic lava]

yea?

Howd they run

its 20gb vram

i thinkn im ok on needing less ram

[ornate elk]

true ^^

[mystic lava]

but the old model takes forever to run

idk why

[ornate elk]

but they are the same speed as the flux fp8 16gb

[mystic lava]

idk why it runs so slow wit flux for me

is it available for the other ui yet

stable

[ornate elk]

only on forge and comfyui

[mystic lava]

ah

[ornate elk]

how slow?

and do you used forge or comfy?

[mystic lava]

comfy

[mystic lava]

like 10 mins to generate slow

[ornate elk]

okay thats to long

[mystic lava]

yeaaaaaa

it should take like 30 seconds lol

tops

[ornate elk]

and you use the 16gb fp8 model?

[mystic lava]

Flux1-dev-fp8

[ornate elk]

check the file size

[mystic lava]

16gb

very strange

that its so sllow

workflow iu use is thbe one u sent me lo

Any ideas @ornate elk

If u can get the speed up

theres a coffee in it for u

💯

[ornate elk]

make sure you dont run anything in the background (games or wallpaper engine)

[mystic lava]

i dont

[ornate elk]

how much RAM do you have?

[mystic lava]

32gb ddr5

😭

[ornate elk]

same as me

[mystic lava]

yea

what are ur gen times

[ornate elk]

50-80 seconds

[mystic lava]

Gyko

gulp

[ornate elk]

try edit the start-comfyui.bat
at the bottom after the --auto-launch you add --reserve-vram 4
then relaunch and test again

[mystic lava]

this ? @ornate elk

[ornate elk]

behind --auto-launch

[mystic lava]

got it

doesnt start now

😭

@ornate elk

[ornate elk]

oh that means your comfyui version is not updated and your using the old torch version

[mystic lava]

how do i go about updating?

[ornate elk]

go into the comfyui folder then click into the file epxlorer bar (not searchbar)
type cmd and hit enter
in the cmd type
git pull
and hit enter again.

[mystic lava]

done

ran

now?

run comf=y

[ornate elk]

yep

[mystic lava]

running

shall rreport speed

after

[ornate elk]

there where a lot of improvements so maybe its faster now

keep watching your taskmanager and the vram usage

[mystic lava]

log

Import times for custom nodes:
   0.0 seconds: C:\Comfy\ComfyUI\custom_nodes\websocket_image_save.py

Starting server

To see the GUI go to: http://127.0.0.1:8188
fatal: No names found, cannot describe anything.
Failed to get ComfyUI version: Command '['git', 'describe', '--tags']' returned non-zero exit status 128.
got prompt
model weight dtype torch.float8_e4m3fn, manual cast: torch.bfloat16
model_type FLUX
Using pytorch attention in VAE
Using pytorch attention in VAE
C:\Comfy\ComfyUI\venv\lib\site-packages\transformers\tokenization_utils_base.py:1601: FutureWarning: `clean_up_tokenization_spaces` was not set. It will be set to `True` by default. This behavior will be depracted in transformers v4.45, and will be then set to `False` by default. For more details check this issue: https://github.com/huggingface/transformers/issues/31884
  warnings.warn(
loaded straight to GPU
Requested to load Flux
Loading 1 new model
loaded completely 0.0 11350.048889160156 True
Requested to load FluxClipModel_
Loading 1 new model
loaded completely 0.0 4777.53759765625 True
C:\Comfy\ComfyUI\comfy\ldm\modules\attention.py:407: UserWarning: 1Torch was not compiled with flash attention. (Triggered internally at ..\aten\src\ATen\native\transformers\cuda\sdp_utils.cpp:455.)
  out = torch.nn.functional.scaled_dot_product_attention(q, k, v, attn_mask=mask, dropout_p=0.0, is_causal=False)
loaded completely 12287.395475097655 11350.048889160156 True
 10%|████████▎                                                                          | 2/20 [00:07<01:16,  4.25s/it]

seems to be going faster tho

at 30 percent already

good sign

the sigma is right

as usual

check kofi

is there any more issues u see in that tho

170 seconds

not the fastest

but faster

what do u think is slowing it down that extra bit

[ornate elk]

idk, you need to have a look at your task manager while generating

ram and vram shouldnt overflow

[mystic lava]

wtf

gpu

is at like

2 percent

utilisation

spikes of 29 percent

but not fully using

[ornate elk]

check the vram usage not the utilisation

dedicated gpu memory

[mystic lava]

generated this in 165 seconds

lemme see vram

[ornate elk]

thanks man!

[mystic lava]

npnp

[azure magnet]

Any help

[ornate elk]

thats normal, got that too

[mystic lava]

its using 18.7gb iof vram

@ornate elk

[ornate elk]

seems okay. you could try change the 4 to a 3 on the --reserve-vram

[mystic lava]

my generation time has doubled

wtf

[ornate elk]

first gen always takes a bit longer

[mystic lava]

i havent changed it

😭

its the third gen

[azure magnet]

So why i get this

[ornate elk]

your on AMD or nvidia?

[azure magnet]

Nvidia

[ornate elk]

make surey our node and comfyui is updated via the manager

[mystic lava]

its at 18 its

but slower

wtf

[ornate elk]

can you screenshot it?

[mystic lava]

we finishing next week wit this one

prompt
Generate a surreal, immersive visual of raw power manifesting as vivid hallucinations. The scene should feel otherworldly, with swirling, abstract shapes and energy waves blending into intricate patterns. These patterns morph and shift, creating a dreamlike landscape where reality bends and fractures. Electric colors like deep purples, radiant golds, and neon blues pulse and glow, casting an aura of intense energy. Figures and symbols from various mythologies and cultures emerge briefly from the chaos, only to dissolve back into the swirling void. The overall mood should evoke a sense of overwhelming, mind-bending power, where the viewer is drawn into a hallucinatory state of awe and confusion.

[ornate elk]

now check the vram

while its slow

[mystic lava]

18/8

gb

18.8

[ornate elk]

hmm

[azure magnet]

Could be this that give errors?

[ornate elk]

hmm idk

[mystic lava]

image finished

[ornate elk]

nice

[ornate elk]

what you could try is to upgrade torch to 2.3.1

[mystic lava]

Import times for custom nodes:
   0.0 seconds: C:\Comfy\ComfyUI\custom_nodes\websocket_image_save.py

Starting server

To see the GUI go to: http://127.0.0.1:8188
fatal: No names found, cannot describe anything.
Failed to get ComfyUI version: Command '['git', 'describe', '--tags']' returned non-zero exit status 128.
got prompt
model weight dtype torch.float8_e4m3fn, manual cast: torch.bfloat16
model_type FLUX
Using pytorch attention in VAE
Using pytorch attention in VAE
C:\Comfy\ComfyUI\venv\lib\site-packages\transformers\tokenization_utils_base.py:1601: FutureWarning: `clean_up_tokenization_spaces` was not set. It will be set to `True` by default. This behavior will be depracted in transformers v4.45, and will be then set to `False` by default. For more details check this issue: https://github.com/huggingface/transformers/issues/31884
  warnings.warn(
loaded straight to GPU
Requested to load Flux
Loading 1 new model
loaded completely 0.0 11350.048889160156 True
Requested to load FluxClipModel_
Loading 1 new model
loaded completely 0.0 4777.53759765625 True
C:\Comfy\ComfyUI\comfy\ldm\modules\attention.py:407: UserWarning: 1Torch was not compiled with flash attention. (Triggered internally at ..\aten\src\ATen\native\transformers\cuda\sdp_utils.cpp:455.)
  out = torch.nn.functional.scaled_dot_product_attention(q, k, v, attn_mask=mask, dropout_p=0.0, is_causal=False)
loaded completely 12287.395475097655 11350.048889160156 True
100%|██████████████████████████████████████████████████████████████████████████████████| 20/20 [02:10<00:00,  6.53s/it]
Requested to load AutoencodingEngine
Loading 1 new model
loaded completely 0.0 159.87335777282715 True
Prompt executed in 170.15 seconds
got prompt
Requested to load FluxClipModel_
Loading 1 new model
loaded completely 0.0 4777.53759765625 True
loaded completely 12274.915006347655 11350.048889160156 True
100%|██████████████████████████████████████████████████████████████████████████████████| 20/20 [02:29<00:00,  7.49s/it]
Requested to load AutoencodingEngine
Loading 1 new model
loaded completely 0.0 159.87335777282715 True
Prompt executed in 165.20 seconds
got prompt
Requested to load FluxClipModel_
Loading 1 new model
loaded completely 0.0 4777.53759765625 True
loaded completely 12274.915006347655 11350.048889160156 True
100%|██████████████████████████████████████████████████████████████████████████████████| 20/20 [05:30<00:00, 16.52s/it]
Requested to load AutoencodingEngine
Loading 1 new model
loaded completely 0.0 159.87335777282715 True
Prompt executed in 358.18 seconds

tutorial

pls

[ornate elk]

@mystic lava edit the start-comfyui.bat again and at the torch line adjust the version numbers:

then delete the venv folder and relaunch

[azure magnet]

Do you know how i can fix it?

[ornate elk]

do you have reactor node installed to?

[azure magnet]

Idk what is rector node

[ornate elk]

ok nvm

[ornate elk]

have you updated comfyui and the nodes ?

maybe also ask in #🧣｜comfy-ui or on the comfyui github

[azure magnet]

yes

[ornate elk]

its defintly an issue with the custom node or comfyui itself as it cant load the onnx dll from insightface

[azure magnet]

Ok i'll try using another node

[ornate elk]

or reinstall it

[azure magnet]

Seam this is the problem this: C:\ComfyUI\ComfyUI\comfy\ldm\modules\attention.py:407: UserWarning: 1Torch was not compiled with flash attention. (Trigg
ered internally at C: \actions-runner_work\pytorch\pytorch\builder\windows\pytorch\aten\src\ATen\native\transformers\cud
a\sdp_utils. cpp:555.)
out = torch.nn. functional. scaled_dot_product_attention(q, k, v, attn_mask=mask, dropout_p=0.0, is_causal=False)

[lofty zenith]

Hi there! 👋

I'm seeking a job as a Full Stack Web Developer and UI/UX Designer. 🧑‍💻 I have 8 years of experience in front-end and back-end development, and I specialize in building websites, web services, microservices, and apps from scratch. 😉

If you have any job openings or new ventures arising, please DM me.

Thanks!

[zenith seal]

hello ! i have a little issue when i use sd 1.5 checkpoints

[high smelt]

is there a way to fix this? after it is 100% it stalled there for roughly 30 min each time

[dry portal]

When selecting checkloads to load in the drop down list, some won't load. I noticed when I downloaded these checkpoints they were 6.5gb.

[devout atlas]

Can anyone help me out with this? It runs extremely slowly and I am not very knowledgeable with this whole program's inner workings.

OSError: None is not a local folder and is not a valid model identifier listed on 'https://huggingface.co/models'
If this is a private repository, make sure to pass a token having permission to this repo either by logging in with huggingface-cli login or by passing token=<your_token>

I downloaded a model from Civiai and installed in the correct folder

[ruby yoke]

how are you trying to use the model? what interface?

[devout atlas]

Like I said, I am a knuckle dragging ape when it comes to this stuff. I am using A1111 with an AMD card

Is that an answer?

[ruby yoke]

that is the correct answer, yes. are you able to use A1111 for other models with no errors?

[devout atlas]

No models work. Although I have only attempted it with 2

[ruby yoke]

alright, to start with, did you go through the AMD install guides linked to this channel?

[devout atlas]

I used the one associated with the A1111 guide on github.

Not here

[ruby yoke]

i'm not familiar with that guide. i know that the ones @ornate elk wrote and that are linked here work. you might look through them.

[devout atlas]

Any one in particular?

Direct ML or Zluda?

[ruby yoke]

you might want to wait till he's back online, actually.

[devout atlas]

Ok would he be ok with me sending him a dm?

[ruby yoke]

i don't know, but when he's online, he's in here and he answers every question asked. he shouldn't be offline for too many more hours

[devout atlas]

Ok well I gtg to bed. So hopefully I will see his response in the morning

[versed imp]

I need a little help about fix my Lora face please

[ruby yoke]

what's wrong with it?

[versed imp]

The face is all deformed and I don't know how to fix it

[ruby yoke]

what do the images in your data set look like, and what model is the lora for?

[versed imp]

Here is the print

[ruby yoke]

is this a lora you trained or just one you're trying to use?

[versed imp]

Is just a Lora that im trying to use

[ruby yoke]

okay, to me i looks like you are trying to use an aspect ratio it doesn't really like. can you make your image height larger than your width and then try generating?

[versed imp]

512x768

[ruby yoke]

your screen shot says it's set to 512x512 - try setting it to 720 height and 576 width

[versed imp]

Yeah i forgot to change sorry, this one is 512x768

[ruby yoke]

that's not so scrunched at least

[versed imp]

Just a litte better but still is deformed

[ruby yoke]

i would personally get rid of everything in the negative prompt box and leave it totally blank

[versed imp]

lets see

Almost same thing

[ruby yoke]

i'm not seeing what I consider deformed - can you explain what you see in the face that isn't correct?

[versed imp]

Yes the face isnt correct

[ruby yoke]

to me, it looks correct. can you please explain what it is about the face that you feel is wrong

[versed imp]

Use a zoon in the face and u ll see

The eyes is wrong, the face is strange

[ruby yoke]

what model is this lora for?

[versed imp]

Do u wanna know the name of model?

[ruby yoke]

i'm guessing this is a pony lora, based on your prompt structure. correct?

[solemn moon]

so i have this error, on startup but, everything works fine..

fatal: not a git repository (or any of the parent directories): .git
venv "C:\Users\o0\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 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]

[ruby yoke]

which interface?

[solemn moon]

auto

[ruby yoke]

maybe recreate your venv

[solemn moon]

im guessing i shld delete venv

yeAa

[ruby yoke]

yeah. try remaking it

[solemn moon]

okei thanks you wizard of crystals

[ruby yoke]

welcome. let me know if you still have issues or not please

[ornate elk]

What resolution did you used?

[ornate elk]

What does the cmd shows ? Did you used upscaling?

[ornate elk]

If they dont load you may need to increase the Windows Pagefile.

[ornate elk]

If your GPU is supported use ZLUDA

[ruby yoke]

morning @ornate elk

[ornate elk]

Morning 👋

[ruby yoke]

that last guy said he'd be back in his morning and wanted to know if he could DM you - do you allow DMs?

[ornate elk]

Yes I do

[ruby yoke]

okay. i didn't know 🙂

[ornate elk]

Np, but I like answer here primarily before going into dm

[ruby yoke]

totally understand that, i do too

[dry portal]

Okay, cause other checkpoints will load (but they're about 2-4gb in size). So increase page file?

[ornate elk]

If you have 16gb RAM then yes

[dry portal]

I have 32gb of system ram.

[ornate elk]

Hmm then it should work

Afk

[ornate elk]

And what's your GPU?

[dry portal]

4070

[solemn moon]

still:

fatal: not a git repository (or any of the parent directories): .git
venv "C:\Users\o0\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 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Version: 1.10.1
Commit hash: <none>
Installing requirements
CUDA 12.1
Launching Web UI with arguments: --xformers --precision full --no-half --lowvram --opt-split-attention --always-batch-cond-uncond --api --ckpt-dir 'D:\stable-diffusion-webui-master\models\Stable-diffusion' --controlnet-dir 'D:\stable-diffusion-webui-master\models\Control Net'

still works though o0

[ruby yoke]

@ornate elk any ideas?

[solemn moon]

startup for auto, but still works..

my .bat

@echo off

set PYTHON="C:\Users\o0\AppData\Local\Programs\Python\Python310\python.exe"
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --xformers --precision full --no-half --lowvram --opt-split-attention --always-batch-cond-uncond --api --ckpt-dir "D:\stable-diffusion-webui-master\models\Stable-diffusion" --controlnet-dir "D:\stable-diffusion-webui-master\models\Control Net"
git pull
call webui.bat

[bold berry]

Hello, my generation with audio file stuck on "pending" stage for 15 minutes already, can someone help?

[zenith seal]

I used 512x1024

[vocal burrow]

you didn't install the webui properly / using the git clone command, did you.

You used the big green "'download as zip" button

[solemn moon]

YESSIR

[vocal burrow]

then that's why git complain, cause you're trying to "update" a regular folder. Not a git one.

it lacks the .git folder

Open a cmd/powershell/gitbash console, go to StableDiffusionWebui's directory and run :

git init
git remote add origin https://github.com/AUTOMATIC1111/stable-diffusion-webui.git
git fetch --all
git reset --hard origin/master
git branch --set-upstream-to=origin/master master
git pull

This should fix git stuff and update SD to the last version.

[solemn moon]

thanking you

[vocal burrow]

and don't use the big green button next time

[solemn moon]

noted

[solemn moon]

getting this:

There is no tracking information for the current branch.
Please specify which branch you want to merge with.
See git-pull(1) for details.

git pull <remote> <branch>

If you wish to set tracking information for this branch you can do so with:

git branch --set-upstream-to=origin/<branch> main

[vocal burrow]

git branch --set-upstream-to=origin/master master

hmmm it's already in the paragraph I've sent you

weird then

git branch --set-upstream-to=origin/master main then ?

[ornate elk]

What does the cmd shows when you try to load the sdxl model?

[solemn moon]

i deleted and busy doing the git clone route

apologies for not even reading the installation instructions properly again 😅

[shut tinsel]

Hello, I am following the following document for finetuning stable-diffusion-xl-1024-v1-0

https://next.platform.stability.ai/docs/features/fine-tuning

in the imports

from stability_sdk.finetune import (
create_model, delete_model, get_model, list_models, resubmit_model, update_model,
FineTuneMode, FineTuneParameters, FineTuneStatus
)
I get the error:

ModuleNotFoundError: No module named 'stability_sdk.finetune'

Can anyone please help me with the fix?

[high smelt]

i used upscaling and cmd shows that it is 100% but i cant find the grid or images if i interupt it or skip it, so i have to wait for 30- an hour for it to be done at 100%

[ornate elk]

Make sure you have set Hires Steps to 10

If not it can take ages

[high smelt]

what does hire steps do? default it is 0

[ornate elk]

Hires steps is the thing that upscales the image.
If its set to 0 it will use the same amount as normal generation steps.
This requires a lot of vram and is not recommend for GPUs below 4080.
So set it to 10 or 15 and it shouldn't freeze anymore.

And its much faster then

[high smelt]

i see, thanks

[dry portal]

@ornate elk in CMD it says runetime error with a bunch of lines saying size mismatch.

[mystic lava]

@ornate elk I updated torch and stuff yesterday like u said

then got this error when i waent to gen something wit flux

very not sigma

[mystic lava]

cs on vacation

well deserved

nice to see

[zenith seal]

hello did you see my reply?

[zenith seal]

yee

[ornate elk]

Can you show it?

[ornate elk]

More like busy week xD

Can check the log later

[quasi ginkgo]

Hey yall, anyone have issues in Forge with all loras showing up when "Always show all networks on the lora page" is disabled?

It also seems like half the defaults I set don't stay set when the UI reloads

[ornate elk]

What was your GPU again?

Can you show the txt2img settings you used?

[zenith seal]

6800xt

basicaly this with no hiighres fix

and res at 1024x768

[ornate elk]

but hires fix is selected in this image

[zenith seal]

yeah im on xl rn

genning stuff

[ornate elk]

and hires steps is set to 0 so it will run out of vram

always set hires steps to 10

[zenith seal]

no but i diin't use high res on sd 1..5

[ornate elk]

and if you use Tiled VAE extension you can alos upscale easily by 2

[ornate elk]

send me an screenshot of the settings if it happens again

[zenith seal]

it happens only when i use sd 1.5 checkpoints

and whats Tiled VAE?

[ornate elk]

can you show it again?

[ornate elk]

its an extension that helps not getting out of memory errors when upscaling at larger resolutions. it also speeds up the generation process a bit

[zenith seal]

basicaly this is the parameters

[loud pond]

hello! Last time i was here i was told i might be able to face swap generations? How would i do this?

[ornate elk]

if you run it now with 1.5 will it run out of memory? can you try

[ornate elk]

hey, thats possible with either the Reactor extension or controlnet with IP Adapter

[zenith seal]

yeah i can

[loud pond]

oh alright thanks

[zenith seal]

how do i use it ?

[ornate elk]

just enable Tiled VAE

dont enable the others

thats all ^^

[zenith seal]

oh yeah i did try with titled and it messup hard

lol

still cool

[zenith seal]

Thanks a lot man you are the best !

[ornate elk]

np 🙂

[zenith seal]

well

[zenith seal]

oh

high res step mb

[ruby yoke]

how much swap space do you have?

[zenith seal]

swap?

[ruby yoke]

paging space.

[zenith seal]

on drive or gpu?

[ruby yoke]

on system

[zenith seal]

8 gb

19 mb

[ruby yoke]

first of all, that is free space, not swap space, and that's not enough to even run your system very well - much less running any applications on it. free up A LOT of space on that drive. put your models on a different drive if you can

[mystic lava]

me?

[mystic lava]

tty

[zenith seal]

well idk what swap space or paging space

[ruby yoke]

swap/paging space is a dedicated area of your drive that the system uses to write temporary files to when there's no space in memory. most people have theirs set by the sytem and it tends to set it as low as possible. sometimes you have to set it higher than the system wants

but in your case, your drive is too full, you need to free up drive space

[zenith seal]

and yeah idk how to free space i have so much porn on my drive

(jk)

but idk what can take all this place

i alrready uninstaled all my game from c drive

[ruby yoke]

external SSD drives aren't that expensive. consider getting one and then putting your models on it, and then just telling your interface where to find them instead of putting them on C:

[zenith seal]

i have a lot of place on d

[ruby yoke]

lots of space free on D: so put your models there and just tell the interface where they are

[mystic lava]

buy 2tb sssd

100 gbp

ish

eaz

[short cove]

Can anyone help me set up an extension? I have ebsynth installed but the ui tab won't show up

[zenith seal]

yeah i can do that right

[ruby yoke]

also - you really shouldn't have anything on your C: drive other than your operating system. you have a very small C: drive. so consider installing games on something else. you might need to pick up a couple more SSD drives or something

[zenith seal]

i don't have any games on c drive realy don't know what that ta muuch space

[ruby yoke]

i know, you said you uninstalled them. i'm saying that if you decide to reinstall something, anything, don't do it on C: get everything off C: other than your operating system and the parts of applications that refuse to install anywhere else. Auto1111 MIGHT need to be installed on C:, but most of what it uses can live somewhere else (example)

[zenith seal]

Yeah i understand that thanks also i gonna move my models on d i think this only iis 200 gb so

wiill be a lot better

just don't see where to change this ::

[ruby yoke]

you have to edit the .bat file i'm pretty sure. @ornate elk should know where

[zenith fiber]

why is it when i generate an image it ussually takes about 10 min but sometimes it takes like 1
with the same parameters
i dont understand it

[zenith seal]

alright bet thanks for your kind help !

[ruby yoke]

any errors in the console log?

[zenith fiber]

how do i open that?

[ruby yoke]

what interface are you using to generate with?

[zenith fiber]

stable diffusion web ui

[zenith fiber]

same as this one

[ruby yoke]

if you move the web UI over and look behind it, do you not see a console window with text?

[zenith fiber]

yeah

[ruby yoke]

that text is your console log

[zenith seal]

wdym

[ruby yoke]

he's just saying he's using the same interface you are

[zenith fiber]

yeah sorry bout that

looks like its just faster suddenly

[ruby yoke]

okay, that just shows stuff is progressing fine. and you're not switching models or anything, are you?

[zenith fiber]

no

i had this before and i saved al the settings

[ruby yoke]

then it might be exactly what @fair oxide said - your filling your VRAM up - and it's using your Ram/CPU instead of your vram/gpu

[zenith fiber]

thats weird

[ruby yoke]

what hardware do you have?

[zenith fiber]

rtx3060 ryzen 9500 32GB ram

[ruby yoke]

ram? or Vram?

[zenith fiber]

ram

that be alot of vram XD

[ruby yoke]

so... you need vram

how much of that do you have?

[ruby yoke]

not for some of the models 😉

[zenith fiber]

hm

[ruby yoke]

and - you are using AMD, and it loves to try to take over and be the GPU instead of letting your GPU be the GPU. please access the pinned posts in this channel and go through all of @ornate elk 's guides

[zenith fiber]

alright thanks

[ruby yoke]

welcome

[zenith fiber]

Yeah cus now it uses al 6gb of vram

Im not sure if it did that before

[ornate elk]

set hires step to 10

never use 0

[cosmic furnace]

Cs1o?;o

[ornate elk]

what txt2img settings you used?

[ornate elk]

hey

[cosmic furnace]

#🤝｜tech-support message Can comment smthing bout zis one?)

Hoi 🙂

Ive heard comfy faster than auto ? Is it?

[oak jay]

is it possible to use Dynamic Thresholding with Flux models on SD Forge? If so, does anyone have any recommended settings please?

[ornate elk]

Auto1111 for extensions, comfyui and forge for flux.
In general Auto1111 is the best with Zluda

[ruby yoke]

it's friendlier at least

[ornate elk]

whats in your webui-user.bat?

[cosmic furnace]

And what from all this is fastest & simplest to get used to ?)

[ornate elk]

Auto

and for flux, forge is easier to use than comfy

[cosmic furnace]

R u sure auto generating images faster ? o.0

[ornate elk]

you said fast and easy

[cosmic furnace]

so forge is better than comfy?)

[ornate elk]

forge could be a bit faster

[cosmic furnace]

And who is fastest of them all?)

[ornate elk]

maybe forge

comfyui has the same speed as auto for me

[ornate elk]

you have to replace the 2 files in comfyui venv lib site packages torch lib again

[zenith fiber]

Just 1024x1024 25 steps and dpm++2m

[ornate elk]

okay and whats inside your webui-user.bat?

[zenith fiber]

Idk what you mean by that

When I click it it just opens the webui

[ornate elk]

right click the file and click edit

there you need to add --xformers --medvram --no-half-vae to the line
set COMMANDLINE_ARGS=
then save and relaunch

[zenith fiber]

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=

call webui.bat

what does that do exactly?

[ornate elk]

like this

that will speed up the image generation and uses less vram

[zenith fiber]

ah alright thanks

[ornate elk]

np, then you shouldnt get any performance issues when using sdxl models for example

[dusky temple]

Seeking Open Source AI for Creating Talking Head Videos from Audio & Image Inputs
The goal of the service is to provide an audio and image of a character, and it generates videos with head movements and lip-syncing.
I know of these open-source models,
https://github.com/OpenTalker/SadTalker
https://github.com/TMElyralab/MuseTalk
but unfortunately, the current output quality doesn't meet my needs.
are there any other tools i didn't know of?
thanks.

[ruby yoke]

there's been some papers released recently, but they're still papers.

[dusky temple]

thats the way with AI stuff huh

[ruby yoke]

it's not easy tech. getting a mouth to move, not that hard. getting the entire face to move and the person to look natural, extremely hard. you'll find the cutting edge for all of this on twitter as that's where the ML scientific community is and where they post first. You might take a look at this video tim put out https://youtu.be/3pp7qw19nuA?si=bO-559HyzA-2wVyD on a new faceswap tool you can run local

[dusky temple]

yeah, you are absolutely right the twitter is the way to go for the AI stuff
thanks for the youtube video, appreciates the efforts

[ruby yoke]

welcome. wish i had better answer

[dusky temple]

Your help means a lot, thanks for trying

[devout atlas]

Hey I am back and trying to install SD on my system which runs AMD. I am following the guide in the pinned tab here. I am very worried about step number 4.

Add the C:\ZLUDA folder and %HIP_PATH%bin to your PATH in the System Variables. Like shown here: (link)

It says if I do this wrong it could brick my system.

[ruby yoke]

you might want to wait till @ornate elk is back online in this case

[fair oxide]

i just got a 4060ti 16gb vram what Args should i put for forge and also for comfy ui? or what is reccomended

[solemn moon]

check pinned messages

[dry portal]

@ornate elk

[mighty echo]

Help?

OutOfMemoryError: CUDA out of memory. Tried to allocate 4.54 GiB. GPU 0 has a total capacty of 8.00 GiB of which 0 bytes is free. Of the allocated memory 11.37 GiB is allocated by PyTorch, and 288.05 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
Time taken: 11 min. 18.3 sec.

[fair oxide]

thank you ^^

[fair oxide]

how do i allocate more memory im having errors trying to use cogvideox I2V i have 16gb vram

[ornate elk]

Edit the webui-user.bat, at the line commandline_args= you need to add: --xformers --medvram-sdxl --no-half-vae
Then save and relaunch

[ornate elk]

Would need the full log + the settings you tried

[ornate elk]

Hey, let me know when you want to try it.

[sterile garnet]

I don't know where else to look but I was wondering if anyone else is up because I'm having an issue thats killing me 😭

I'm having a problem with flux on comfy UI and I'm honestly starting to believe I've screwed it beyond repair
I was getting 52 it/s (and terrible outputs, had it to 20 steps and 512x512 but it kept coming out looking like a 4 step render) and generation times were taking a good 23 minutes with a 3050ti and i7 laptop and from my task manager it looks like I have enough but I was getting terrible outputs and switched to dev NF4.

But then I heard downgrading to CUDA 12.1 may help, so I did that.
And now whenever I try to process an image I get this error popup

I also noticed its only using my dedicated ram and not shared GPU

anyone have any idea on how to resolve this?

[shut tinsel]

anyone fine tuned SDXL thorugh stability ai's API?

[tiny perch]

Hey, I have a question regarding the commercial usage of the latest api. If I want to build and deploy an app using the API but don't plan on monetizing it, does it count as non-commercial?

[mystic lava]

which ones sorry?

[ornate elk]

Checkout my Comfyui Zluda guide at step 5

[wicked edge]

anybody helps me explain why this always error

[ornate elk]

What does the cmd shows?

[wicked edge]

just error

[ornate elk]

Thats not the cmd (console)

[dry portal]

@ornate elk

[mystic lava]

aight donme

just installed win 11

forced me to update

we'll see howe it goes

Starting server

To see the GUI go to: http://127.0.0.1:8188
fatal: No names found, cannot describe anything.
Failed to get ComfyUI version: Command '['git', 'describe', '--tags']' returned non-zero exit status 128.
got prompt
got prompt
model weight dtype torch.float8_e4m3fn, manual cast: torch.bfloat16
model_type FLUX
Using pytorch attention in VAE
Using pytorch attention in VAE
C:\Comfy\ComfyUI\venv\lib\site-packages\transformers\tokenization_utils_base.py:1617: FutureWarning: clean_up_tokenization_spaces was not set. It will be set to True by default. This behavior will be deprecated in transformers v4.45, and will be then set to False by default. For more details check this issue: https://github.com/huggingface/transformers/issues/31884
warnings.warn(
Requested to load FluxClipModel_
Loading 1 new model
loaded completely 0.0 4777.53759765625 True

RUNNING RN

@ornate elk

gyulp

[ornate elk]

Your webui-user.bat looks incorrect and also you tried to install dreambooth which is known for making stable diffusion issues.
So reset your webui-user.bat to defaults. Delete the dreambooth extension and delete the venv folder.
Then relaunch the webui-user.bat

[dry portal]

How do I reset the user bat to default?

[ornate elk]

whats your gpu?

[dry portal]

4070

[ornate elk]

okay then delete everything inside the webui-user.bat and replace it with this:
`@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--xformers --no-half-vae
git pull
call webui.bat`

[dry portal]

Sounds good and then also delete dreambooth and venv folder from SD?

[ornate elk]

you have the same issue as before. you need to replace the cublas and cusparse files in the venv

[mystic lava]

ComfyUI Error Report

Error Details

• Node Type: CLIPTextEncode
• Exception Type: RuntimeError
• Exception Message: CUDA error: CUBLAS_STATUS_NOT_SUPPORTED when calling cublasSgemm( handle, opa, opb, m, n, k, &alpha, a, lda, b, ldb, &beta, c, ldc)

Stack Trace

  File "C:\Comfy\ComfyUI\execution.py", line 323, in execute
    output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)

  File "C:\Comfy\ComfyUI\execution.py", line 198, in get_output_data
    return_values = _map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)

  File "C:\Comfy\ComfyUI\execution.py", line 169, in _map_node_over_list
    process_inputs(input_dict, i)

  File "C:\Comfy\ComfyUI\execution.py", line 158, in process_inputs
    results.append(getattr(obj, func)(**inputs))

  File "C:\Comfy\ComfyUI\nodes.py", line 65, in encode
    output = clip.encode_from_tokens(tokens, return_pooled=True, return_dict=True)

  File "C:\Comfy\ComfyUI\comfy\sd.py", line 126, in encode_from_tokens
    o = self.cond_stage_model.encode_token_weights(tokens)

[dry portal]

@ornate elk The CMD was downloading a bunch of stuff, but anyway tried loading those checkpoints and it works fine now.

[ornate elk]

show me you C:Zluda folder

[dry portal]

Thanks so much for the help.

[ornate elk]

no problem 🙂

[ornate elk]

if you want to train loras you should use sperate tools like Kohya_ss or OneTrainer (instead of dreambooth)
They dont break your Stable diffusion as they are no extensions

[mystic lava]

I had deleted the non 64 veresion

[ornate elk]

Why?

[mystic lava]

Uhhh

[ornate elk]

You only need to move the 64_11 DLL files into the lib of torch in Comfyui and overwrite

[mystic lava]

wait what

[ornate elk]

Go into the C:\ZLUDA\ Folder. There copy the cublas.dll and the cusparse.dll and repaste them inside the folder again. Rename the copys to cublas64_11.dll and cusparse64_11.dll Copy these two files to ComfyUI\venv\Lib\site-packages\torch\lib and overwrite if asked.

[mystic lava]

It now just crashes

😭

[ornate elk]

Any error?

[mystic lava]

closes befoire i can see anything

[ornate elk]

is your environment path correct?

[mystic lava]

i think so

[ornate elk]

check it

you said you installed Win11

maybe it has reset some stuff

[ornate elk]

looks good

[mystic lava]

is there a crash log i can senbd u

[ornate elk]

nope

where is your comfyui folder located?

[mystic lava]

C:

[ornate elk]

full path please

[mystic lava]

C:\Comfy\ComfyUI

[ornate elk]

ok

and when did comfyui start to not work?

is your auto1111 still working?

[mystic lava]

it started not wokring

after i updated stuiff

[mystic lava]

like non flux?

[ornate elk]

stuff?

[mystic lava]

yea

torch i think

or like comfy itslef

[ornate elk]

the other webui, if you still have it

[ornate elk]

okay then delete the venv folder.
relaunch comfyui and then wait until its done.
then replace the 2 files in torch/lib again and restart

[mystic lava]

delete venv

relaunch

replace filers

got it

done

still crashing

wtflip

dude

[ornate elk]

maybe you need a reinstall then

[mystic lava]

ffs

😭

shouyld just bought a 3080

[ornate elk]

have you installed Automatic1111 with zluda before?

[mystic lava]

i t hink ive already done oit

[ornate elk]

try if that still works

[mystic lava]

il do later

gtg for a biot

tysm

[ornate elk]

alright np

[gusty salmon]

why does automatic1111 always try to flip people if the face isnt facing the camera? i feel like generating the back of a head is so hard

[uncut totem]

any idea when using upscaler getting this error.

TypeError: 'float' object cannot be interpreted as an integer

[ornate elk]

Can you show your txt2img settings?

[uncut totem]

let me run again to give you a good post

[ornate elk]

Is the lora a flux lora?

Does upscaling sdxl works?

[uncut totem]

flux lora yes

ill try sdxl

[devout atlas]

I’ll be home in around 5 hours. I’m unaware of your schedule

[ornate elk]

Thats pretty late for me but maybe I'm still on

[devout atlas]

Oh weeeellll tomorrow is Saturday. Maybe try then? I don’t want to keep you up

[ornate elk]

No all good. Just @ me when your ready. If I have time I can help

The Path setting isnt something that takes long

[devout atlas]

Ok I’ll ping you when I’m home

[devout atlas]

@ornate elk You around?

[ornate elk]

Yes

[devout atlas]

Okie. Can we do the step #4

[ornate elk]

Sure

[devout atlas]

Can I dm and screen share with you?

Make sure I don’t melt my Pc

[ornate elk]

Give me 5 mins

[devout atlas]

Ok

[ornate elk]

ok im ready

[whole anvil]

I want to try to train my own lora with my original characters so i can generate stuff with then as opposes to trying to recreate them with different stuff. Is this possible and how difficult would this be for a beginner?

[devout atlas]

Thank you to @ornate elk for being an absolute alpha chad and helping me get SD to run on my AMD card. You rock brother.

[ornate elk]

No problem ^^ was fun!

[rugged latch]

any one um

uh. anyone got a cool css i can "borrow"

i got one but, i jus made it change the colour of the generate button and sliders

i wan sum like, wit animations and like, idk, sounds maybe

[tulip gust]

Guys I need help. I've been trying to get stable diffusion working on my PC but im using a 6900XT from AMD
and everytime I attempt to generate an image it just tells me that no NVIDA drivers were detected 😦

[terse siren]

Is there any way to disable your GPU fan

because I hate how loud mine gets when I use SD

Fan Control doesn't do shit

[tulip gust]

its not something I would recommend

I'm new to SD but I know it requires intensive use of the GPU. The fans being loud is it just attempting to cool down the GPU during the intensive work load

you could attempt to clean your gpu and that possibly could help

[ruby yoke]

read through the guides pinned in this channel. look for the push pin icon on the top (if on desktop) or bottom (if on mobile)

[tulip gust]

Thank you, i've possibly made a mistake somewhere.

[terse siren]

I dont intend to leave it disabled permanently

only when I'm using SD

[mighty echo]

Let's try it

[wicked edge]

I have no idea

[fresh current]

Anyone have used https://civitai.com/models/560402/speciosa-25d this model? It just refused to load for me

when I try to load it it give me size mismatch error

[ornate elk]

Hey checkout my AMD install guide from the link in the pinned messages

You want the Automatic1111 with ZLUDA guide. For the best performance/usage

[ornate elk]

Your probably hearing Coil Whine.
Thats not from the Fan.

[ornate elk]

Can you show the full cmd log?

[ornate elk]

Would need to see the full cmd log

[wicked edge]

okie here it is

[ornate elk]

I mean all of it, scroll upwards and mark everything to copy

Or hit Ctrl+A

[wicked edge]

Is it?

[ornate elk]

Okay first remove the --skip-version-check from the webui-user.bat
Then relaunch.
And you can't use openpose 1.5 model with an sdxl model.
For sdxl you need an openpose sdxl model.

[wicked edge]

openpose sdxl model where I can find it?

[ornate elk]

Here;
https://huggingface.co/lllyasviel/sd_control_collection/tree/main

Try the 700mb one from the bottom

[wicked edge]

oh, download to this folder Stable-diffusion\stable-diffusion-webui\extensions\sd-webui-controlnet\models, Is it?

[ornate elk]

For sdxl a native of at least 1024x768 is required

[wicked edge]

ok let me try

[wicked edge]

also cant

or the hardware not support

[noble belfry]

it looks sarong but it feels saright

when you press this, what does it say in the cmd window (the black window with all the writing)

[rugged latch]

BOOT FAILED
the following file is either missing or currupt C:/windows/system/kernal32.dll

please insert a bootable disk and press enter.

[wicked edge]

How to fix ah

[rugged latch]

xD

[ornate elk]

Can you show the cmd log again?

[ornate elk]

Okay, seems like it can't download the preprocessor model because of an SSL error in python.
Try upgrading python to 3.10.11 64bit and if your from Asia you may need to change your DNS or use a VPN

[wicked edge]

lemme try

[warped canyon]

Why is ip adapter face id not working on Forge? It's working on A1111 but it's not working on forge, saying that I need to install insightface, but I have it installed... (I think)

[mortal compass]

What happened to the old tree structure? I used to be able to see all the loras to the left in a folder tree, now it's looking like this 🤔 Is this a me problem or has anyone else noticed this in the new Forge?

[thorn badge]

does anyone know why forge tells me this when trying to load checkpoints?
Traceback (most recent call last):
File "Z:\Forge\stable-diffusion-webui-forge\modules_forge\main_thread.py", line 30, in work
self.result = self.func(self.args, **self.kwargs)
File "Z:\Forge\stable-diffusion-webui-forge\modules\txt2img.py", line 123, in txt2img_function
processed = processing.process_images(p)
File "Z:\Forge\stable-diffusion-webui-forge\modules\processing.py", line 800, in process_images
p.sd_model, just_reloaded = forge_model_reload()
File "Z:\Forge\stable-diffusion-webui-forge\venv\lib\site-packages\torch\utils_contextlib.py", line 115, in decorate_context
return func(args, kwargs)
File "Z:\Forge\stable-diffusion-webui-forge\modules\sd_models.py", line 503, in forge_model_reload
sd_model = forge_loader(state_dict, additional_state_dicts=additional_state_dicts)
File "Z:\Forge\stable-diffusion-webui-forge\venv\lib\site-packages\torch\utils_contextlib.py", line 115, in decorate_context
return func(*args, kwargs)
File "Z:\Forge\stable-diffusion-webui-forge\backend\loader.py", line 272, in forge_loader
raise ValueError('Failed to recognize model type!')
ValueError: Failed to recognize model type!
Failed to recognize model type!
I have put checkpoints in the correct folder

[azure magnet]

Someone know how to solve this: ComfyUI Error Report
Error Details
Node Type: IPAdapterUnifiedLoaderFaceID
Exception Type: Exception
Exception Message: DLL load failed while importing onnx_cpp2py_export:
Stack Trace

  File "C:\ComfyUI\ComfyUI\execution.py", line 323, in execute
    output_data, output_ui, has_subgraph = get_output_data(obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
                                           ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

  File "C:\ComfyUI\ComfyUI\execution.py", line 198, in get_output_data
    return_values = _map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb)
                    ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^

  File "C:\ComfyUI\ComfyUI\execution.py", line 169, in _map_node_over_list
    process_inputs(input_dict, i)

  File "C:\ComfyUI\ComfyUI\execution.py", line 158, in process_inputs
    results.append(getattr(obj, func)(**inputs))
                   ^^^^^^^^^^^^^^^^^^^^^^^^^^^^

  File "C:\ComfyUI\ComfyUI\custom_nodes\ComfyUI_IPAdapter_plus\IPAdapterPlus.py", line 610, in load_models
    self.insightface['model'] = insightface_loader(provider)
                                ^^^^^^^^^^^^^^^^^^^^^^^^^^^^

  File "C:\ComfyUI\ComfyUI\custom_nodes\ComfyUI_IPAdapter_plus\utils.py", line 161, in insightface_loader
    raise Exception(e)

[ruby yoke]

maybe reinstall insightface then

[wraith valley]

does anyone now why i have no encoders there ?

in the video i have watched he used same model

[ornate elk]

your using Automatic1111, and the viode probably used Forge webui

[wraith valley]

https://youtu.be/msb3f0ytDso?si=_z4m1Z3XLG9Eue_7&t=536

its this one i put the same minute in the link

[ornate elk]

he has already downloaded some VAE files

you need to download them and put them into models/vae then they will show up

[wraith valley]

okay i see
he said that the encoders are integrated in the models that confused me

thanks man

[ruby yoke]

VAE can be integrated in a model, but vae isn't the same thing the encoders that you run your prompt into

[wraith valley]

okayyy i have to read more about it thanks

[ornate elk]

he says that the vae are listed on the models page if needed

[wraith valley]

oh okay got it

[median geode]

hey guys, what does it mean?

[ornate elk]

hey, your internet connection was so slow that the download of torch got interrupted

best you can do is plug the PC into Ethernet

[ornate elk]

whats the model?

[wraith valley]

are files with .safetensors in the end always vae files ?

[ornate elk]

.safetensor can be models, vae and upscaler

vae files are always around 330mb

[wraith valley]

hmm okay

https://civitai.com/models/441550?modelVersionId=527934

this one has 70mb

its a model

[ornate elk]

thats a lycoris

[wraith valley]

wtf

[ornate elk]

always check the info box

[wraith valley]

ah okay

[wraith valley]

does someone have mac here ?

[primal pendant]

hmmmm was following the guide exactly for [AMD] Automatic1111 with ZLUDA for my 6700xt but already hada issue after final step

venv "C:\SD-Zluda\stable-diffusion-webui-amdgpu\venv\Scripts\Python.exe"
ROCm Toolkit ROCm was found.
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Version: v1.10.1-amd-10-g2872b02d
Commit hash: 2872b02d3b935665c1b52a32e8bc53b07ec5d540
Failed to install ZLUDA: HTTP Error 404: Not Found
Using CPU-only torch
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
C:\SD-Zluda\stable-diffusion-webui-amdgpu\venv\lib\site-packages\pytorch_lightning\utilities\distributed.py:258: LightningDeprecationWarning: pytorch_lightning.utilities.distributed.rank_zero_only has been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it from pytorch_lightning.utilities instead.
rank_zero_deprecation(
Launching Web UI with arguments:
ONNX: version=1.19.2 provider=CUDAExecutionProvider, available=['TensorrtExecutionProvider', 'CUDAExecutionProvider', 'CPUExecutionProvider']
ZLUDA device failed to pass basic operation test: index=None, device_name=AMD Radeon RX 6700 XT [ZLUDA]
CUDA error: operation not supported
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1.
Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.

[mystic zodiac]

i have a question,
i am using SD model with some SD poses LoRa, but the problem spreads to everything i am using on webui
When i slightly change the prompt to get what i need for a number of times, i am getting almost what i want,
but then i want to get closer to initial images, so i use an exact older prompt,
But i am still getting the same poses/styles/colors i just erased in the prompt
- is there some type of session memory buffer generation helper or something i am not aware of?

[ornate elk]

Failed to install ZLUDA: HTTP Error 404: Not Found
Make sure oyu have zluda in the env path

[primal pendant]

only has to be system right ?

[ornate elk]

yes

[primal pendant]

or should i do it for variables to ?

[ornate elk]

nope its good there

[primal pendant]

still does same though

😭

[ornate elk]

try deleting the .zluda folder inside the stable diffusion webui folder

[primal pendant]

oh will it try to do again ?

nope

i'm cursed when it come to guides i did nothing wrong i swear it

i followed it exactly

[ornate elk]

but you have python 3.10.6 and my guide lists 3.10.11 for example

whats you AMD Adrenalin version?

[south cedar]

does anyone know what this is and why its 10 gigs?

[ornate elk]

it could be the clip model

[south cedar]

ahh i see is there anyway to change the download location of this file?

[primal pendant]

so any ideas?

[ornate elk]

nope

[ornate elk]

have you installed hip sdk and downloaded the correct rocm library files?

[primal pendant]

yeah then rebooted like guide said

[ornate elk]

and have you replaced the library files correctly?

[primal pendant]

yeah it's where library renamed orignal then put newer one

[ornate elk]

try deleting the venv folder and relaunch

[primal pendant]

ok will do

[south cedar]

im sorry but im not famliar with these terms i have a Nvidia card tho if that answers anything

[ornate elk]

that message was for munchie sry xD

[south cedar]

oh lol

[primal pendant]

lol

same

[ornate elk]

http error from zluda is strange

any vpn open ?

or firewall that blocks

[primal pendant]

not using vpn

i will disable and try

[ornate elk]

and third party antivirus programm?

[primal pendant]

only windows security crap

[ornate elk]

ah okay, that shouldnt be an issue

dont need to disable firewall or defender

[primal pendant]

i disabled it all

nothing different

[ornate elk]

what hip sdk version did you installed?

[primal pendant]

the one from guide

as i'm 6700 xt

newer only applied to 7000

or 6800 and upwards

should i jsut delete the other one ?

i;m jsut gonna delete everythign start over

i really don't think i missed anything though

all was fine till final run part

i knew it was going to well lol

[ornate elk]

not needed

[primal pendant]

so any ideas?

i followed the guide exactly

maybe add more here

?

[ornate elk]

nope

[ornate elk]

open up a cmd as admin and run
sfc /scannow
also
pip cache purge

[primal pendant]

ok

all good

reboot run again ?

i'm gonna remove amd rocm to and reinstall '

[ornate elk]

you can also install hip sdk 6.1
and use these rocm library files:
https://github.com/likelovewant/ROCmLibs-for-gfx1103-AMD780M-APU/releases/download/v0.6.1.2/rocm.gfx1031.for.hip.sdk.6.1.2.7z

[primal pendant]

which is morte liekly to work ?

[ornate elk]

both can work but 6.1 has better future support

[primal pendant]

so go for 6.1?

so should i remove stable to ? ?

[ornate elk]

yea better start fresh

[primal pendant]

manual delete enough right ?

or need to run ? something

[ornate elk]

nope as you already cleared the pip cache

[primal pendant]

ok

same version as before besides the rocm ?

or anythign better ?

i'm complete clueless newb right now

all new to me

but guide seemed so easy to follow

shall i remove paths to ?

[ornate elk]

nope

[primal pendant]

ok if it don't work this time i will jsut quit lol

i gave up smoking like few weeks ago this making me wanna start up again

[ornate elk]

nah we can get it wo work

helped a lot of people getting it to run

[primal pendant]

ok that's nice

so same version again ?

besides rocm?

this time 6.1

i was wondering maybe i don't run this thing as fails anyway ?

or does it need to fail ?

nevermind ran it then closed it as guide says

must be reason for it

[ornate elk]

you need to run it

[primal pendant]

yeah thought as much

jsut at zluda part putting file on the c

[ornate elk]

make sure you download the one for rocm6

[primal pendant]

yeah ofc

[ornate elk]

and update your winrar lol xD

there was a big security flaw in the older versions

[primal pendant]

oh this one got something different with it

the version 5 didn't have this

[ornate elk]

yep

[primal pendant]

where do i put this ?

[ornate elk]

leave it there

[primal pendant]

oh ok

[ornate elk]

rocblass dll needs to be in rocblass folder

[primal pendant]

so why did version 5 not have this ?

well reboot time again lol

now i pray 🙏

oh

this is better so far lol

so version 5 was messed up and caused all that lol

that person needs to change the guide

so far so good

[ornate elk]

im that person xD

[primal pendant]

spoke to soon

hahahahha

chnage it then

[ornate elk]

will do

in the next days

[primal pendant]

that version 5 was sobroken

[ornate elk]

whats the model?

[primal pendant]

sec

it said in guide to puta sd 1.5 even in so i dropped the anime one

looked good

ok so now i close it?

[ornate elk]

yep

have you done the zluda stuff?

and restarted the pc ?

[primal pendant]

not yet doing now

oh env you mean yeah all that i was reading wrong dam thing jsut now and was looking at wrong guide scrolled to far lol

cublas.dll

i was lookign at thta's was comfyi

that was lucky

so in zluda i do nothign in way or naming ? things different ?

like the comfyui one says

[ornate elk]

yep

[primal pendant]

ok

so after path and reboot should jsut work as it should ?

sound to easy

something gonna mess up lol

[ornate elk]

yep

[ornate elk]

i see here a big issue.
OneDrive sync

[primal pendant]

onedrive?

[ornate elk]

yea that folder seems like OneDrive synced

[primal pendant]

oh

how i fix that ?

i don't even have onedrive installed

[ornate elk]

okay strange

that green arrow mark

[primal pendant]

i think cause it's fully open c drive all rights

as i thought that was my issue to

so jsut gave all users full

as somethimes permissions mess stuff up didn''t want to take chances

so best way give access to all users full control

god dmann it why

i'm gonna smoke before i break something please help

[ornate elk]

thats okay

just delete the .zluda flder again

[primal pendant]

which one ?

the c root one ?

or this

[ornate elk]

the .zluda

[primal pendant]

ok done

modelfailed

is that bad /

[ornate elk]

can you try an other model?

[primal pendant]

tell me one ?

i'm so cluless comeple newb

there are so many

https://civitai.com/

some are massive

[ornate elk]

https://civitai.com/models/4384/dreamshaper

[primal pendant]

ok dl now

some i saw was like 20 + gb

btw is there a guide on 6700 xt best settings ?

2 mins and it will be done slow dl

[ornate elk]

nope

[primal pendant]

dam

seems hella slow

not good lol

😭

[ornate elk]

can you show me the full cmd log?

[primal pendant]

the whole text ?

i don't think i can do that much i bet discord won't allow

[ornate elk]

you can

as txt file

[primal pendant]

yeah

zluda again i think lol

[ornate elk]

nope

maybe torch

zluda got recognised properly

delete the venv folder and relaunch

[primal pendant]

ok that's good then

ok

[ornate elk]

and delete the base64 model

[primal pendant]

oh

ok

[ornate elk]

looks good

[primal pendant]

so is that model dodgey then like it had alot of downlaods

[ornate elk]

maybe

we will see

[primal pendant]

that's progress

hmmm

[ornate elk]

thats ok

[ornate elk]

looks good

[primal pendant]

well it's cerainly doing more but it's not using my gpu

[ornate elk]

yea compiling uses cpu

this process can take a long time

15-40 mins

[primal pendant]

then will use gpu ?

[ornate elk]

yes for the image generation

[primal pendant]

ok cool thanks for the help i think you saved my pc from taking a few hits llol

[ornate elk]

no problem

[primal pendant]

so whats your set up nvidia i bet ?

[ornate elk]

AMD

[primal pendant]

oh

[ornate elk]

7900XTX

[primal pendant]

very nice