Hopefully a simple question, learning about Loras, I'm currentingly using module illustrij_v825d, what type of loras work with that? Is it all SDXL's? Or only certain ones, or? Cus there's distilled, lgithening, 1.0, 0.9, so many SDXL lora types I don't know which one works with that. Or maybe none of them and I have no idea what I;m doing?
#🤝｜tech-support
turbid willow
turbid willow
I think I figured it out, I think I need ones specifically for Illustrious, that makes sense, doh. XD
ornate elk
I think I figured it out, I think I need ones specifically for Illustrious, that...
Yep first check for Illustrious loras, but also some sdxl and pony loras work with it too
Like add_detail
turbid willow
What is add_detail?
ornate elk
What is add_detail?
A famous lora
https://civitai.com/models/122359/detail-tweaker-xl
turbid willow
Ooh, thanks.
Also, I just turned on ONE lora, and now instead of taking 1 minute to generate an image, it says it will take 20+ minutes. Is that normal!? @_@
That can't be right, right? Sounds like I did something wrong or broke something. In fact, normally it takes less than 1 minute for me to generate an image.
ornate elk
What's your GPU again?
ornate elk
Do you have --xformers in the webui-user.bat?
turbid willow
Idk if I know what that is
turbid willow
Forge web ui
turbid willow
Not sure I know how, but I can try to google that. Are you use that would help? my tast manager says my GPU is already being used 100% by python.
ornate elk
Not sure I know how, but I can try to google that. Are you use that would help? ...
Can you screenshot the webui?
turbid willow
Which part?
ornate elk
Everything
turbid willow
ornate elk
Ahh ok wrong hires fix settings are the issue here
turbid willow
Oh crap... You're right, I normally never use latent
ornate elk
Always set the hires steps to 10.
If its on 0 it will take the same amount as normal steps which would be to much.
Then set the denois to 0.5
turbid willow
Thanks, I'll try that, also that was a totally DOH on my part lol, I don't normally use Latent, I must have forgot to set it correctly this time.
ornate elk
Oh and important is to enable xl mode at the top left
turbid willow
I'll try that,
ornate elk
Its needed for sdxl, illust and pony
turbid willow
thanks! I think I tried it once and it didn't work so I stuck with sd, but I'll try again.
Is there any way to make web ui save ur settings so you don't have to enter them each time you open it?
ornate elk
You can try edit the config_ui.json for that
turbid willow
Oh, cool!
I'll look into that. Man, I learn so much every day it's crazy. Thanks so much ❤️
ornate elk
Easier Is to just press the blue button below the generate button to get the last settings used
turbid willow
Oooh, didn't know that was there. Thanks! lol
ornate elk
Or you drop an image into the positive prompt box and then press the blue button to apply the same settings
ornate elk
I'll look into that. Man, I learn so much every day it's crazy. Thanks so much ❤...
Np, there is a lot to learn and a few tricks too
Most people dont know the can drop an mp3 file into the webuis folder and that sound then gets played everytime an image is generated.
Should be named Notification.mp3
turbid willow
Ha, that's cool!
I need the ff7 combat victory sound effect
Damn... still says 32 minutes eta
I swear, I never had this problem till just a bit ago
ornate elk
Do you run anything in the background like games or Wallpaper engine?
turbid willow
nope
ornate elk
Or twitch, yt etc
ornate elk
Hmm can you screenshot again?
turbid willow
ornate elk
What does the taskmanager GPU performance tab shows?
turbid willow
ornate elk
Try to generate again but this time using 768x1024 as resolution
turbid willow
Sec, I have to close it all down and re-opn it cus the interrupt button never works for me
ornate elk
Also make sure your GPU driver is updated
turbid willow
That's also a good idea, I should do that. But seriously, I didn't have this problem an hour ago
ornate elk
Did you used there hires fix too?
turbid willow
Hmm, that blue button under generate didn't work, shame,
ornate elk
Is your webui updated?
ornate elk
Ok
ornate elk
Do you have multiple displays? Or do you have windoes Wallpaper set to switch backgrounds?
ornate elk
Ok
turbid willow
Took like 10 seconds
Normal for me
I'll add my other prompt back now but with out the lora and test
smooth so far, almost done
adding Lora, trying again.
Yuuuup, it's that goblin lora, don't know why. Now it's saying 30 minute eta
So weird lol
Trying a different lora
ornate elk
Also check the vram usage
In taskmanager
But in the performance tab
And then on GPU directly
turbid willow
this?
btw switched to a different lora, using a norman rockwell style lora, says 3 minutes to make one image. That's a LOT better than the last Lora.
So maybe some Lora's are just really bad?
I will totally try updaing my drivers later today though when I have some free time.
Rip, when I added that Lora, it gave me a black image.
ornate elk
Maybe the lora wasn't compatible
turbid willow
Maybe, I just got some that said they worked with Illustrious.
But maybe my version of illustrious is different
ornate elk
The driver is from august 2024 
turbid willow
Oh, didn't know it made that big of a difference. Wow.
ornate elk
Lots of stuff gets fixed and improved over time.
You can gain more fps in games too
turbid willow
Do you think that might be why these lora's are screwing up?
Or is it some deeper issue?
Like wrong version
This is what I'm using
ornate elk
Its hard to say what caused the massive slow down.
But the broken image from the other lora was due incompatibility mostly.
turbid willow
Thank you so much for all the help ❤️
ornate elk
No problem, let me know when you made the driver update + PC restart.
Then we can check again
turbid willow
It wont be for like 7 hours or so
I need to get ready to leave, gotta go touch grass and stuff
ornate elk
Oh okay ^^
ornate elk
Np, and Thanks! 🙂
turbid willow
I found out i can interrupt generation! I just have to click the button 20 times for it to work.
vast crest
hi, can i ask for help also for config-ui?
ornate elk
I found out i can interrupt generation! I just have to click the button 20 times...
double click is mostly enough, and if you enable never OOM for Tiled VAE at the bottom its even faster and uses less resources while upscaling
ornate elk
hi, can i ask for help also for config-ui?
comfyui? sure, if its technical, also we have a #🧣｜comfy-ui channel
ornate elk
AFK, driving
quick fiber
AFK, driving
alright take your time
fair oxide
I just downloaded and installed (i hope correctly) stable diffusion, is there anything glaringly obvious from this screenshot as to why my images are coming out like this? using a 4060
vocal burrow
I just downloaded and installed (i hope correctly) stable diffusion, is there an...
you're using (the outdated) sd2.1 model which is trained for 768x768 resolution while requesting a 512x512. Also 13.5 is way too high for CFG, lower it to 6 or 7
fair oxide
you're using (the outdated) sd2.1 model which is trained for 768x768 resolution ...
alrighty, should i update to 3.5 or 3? im sure my gpu will fry with sdxl
vocal burrow
follow the guide from the pinned comment. Any 4060 is capable of using sdxl or sd3.5
(sdxl is older than sd3 and sd3.5 btw)
fair oxide
perfect thank you
broken goblet
Anyone knows how can I use this on ForgeUI? https://huggingface.co/zer0int/CLIP-GmP-ViT-L-14 (the text model)
Reason: the text in my gens suck, the model can't do properly the number I prompt, I heard this CLIP makes better text and that is also compatible with SDXL, I use Illustrious so idk if I'm getting this weird gens because of that or because I did something wrong
quick fiber
@ornate elk are u free now?
thorny lodge
hey guys i would need a help of making more realistic ai picture
random night
hey guys i would need a help of making more realistic ai picture
which model did you use?
thorny lodge
which model did you use?
realistic vision
random night
realistic vision
do you know the extension as style selector?
thorny lodge
do you know the extension as style selector?
i used picture girl in bikini and wanted to create a model
and it got really bad
let just say that
random night
i used picture girl in bikini and wanted to create a model
https://github.com/ahgsql/StyleSelectorXL.git try this one
thorny lodge
https://github.com/ahgsql/StyleSelectorXL.git try this one
can i send you in DM's?
a picture
so you can see?
random night
umm, sure
fair oxide
follow the guide from the pinned comment. Any 4060 is capable of using sdxl or s...
I just wanna say thanks for the help, my first art was inspired by your pfp
ornate elk
Anyone knows how can I use this on ForgeUI? https://huggingface.co/zer0int/CLIP-...
You cant use this model to generate images
Its a clip model that has to be used differently
broken goblet
Its a clip model that has to be used differently
I can't use it on Forge?
ornate elk
I can't use it on Forge?
you cant use it ti generate images, but you can use it as clip model if a model like flux needs it
broken goblet
Oh, so the only way is to use it on ComfyUI, the author says it's compatible for SDXL so I guess it works there
ornate elk
Oh, so the only way is to use it on ComfyUI, the author says it's compatible for...
you can try, idk
ornate elk
<@161910921269084160> thats the whole thing
you have a permision isue there
you need to reinstall the webui and then dont launch any .bat as administrator
that messes up the permission
quick fiber
you have a permision isue there
rhis doesnt work for me
ornate elk
rhis doesnt work for me
press enter
quick fiber
press enter
quick fiber
i assume i delete what i had before
ornate elk
i assume i delete what i had before
but you need a folder
quick fiber
yeah
ornate elk
where did you installed it?
random night
i will delete stability matrix right away
ornate elk
you should follow my guide for the install and not install the webui into Appdata
didnt saw it was comfyui so --use-zluda isnt needed
random night
didnt saw it was comfyui so --use-zluda isnt needed
Activating virtual environment...
Running main.py with --auto-launch --disable-xformers --use-quad-cross-attention --reserve-vram arguments...
Checkpoint files will always be loaded safely.
Traceback (most recent call last):
File "C:\AI\ComfyUI\main.py", line 136, in <module>
import execution
File "C:\AI\ComfyUI\execution.py", line 13, in <module>
import nodes
File "C:\AI\ComfyUI\nodes.py", line 22, in <module>
import comfy.diffusers_load
File "C:\AI\ComfyUI\comfy\diffusers_load.py", line 3, in <module>
import comfy.sd
File "C:\AI\ComfyUI\comfy\sd.py", line 6, in <module>
from comfy import model_management
File "C:\AI\ComfyUI\comfy\model_management.py", line 166, in <module>
total_vram = get_total_memory(get_torch_device()) / (1024 * 1024)
File "C:\AI\ComfyUI\comfy\model_management.py", line 129, in get_torch_device
return torch.device(torch.cuda.current_device())
File "C:\Users\nhung\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\cuda\__init__.py", line 878, in current_device
_lazy_init()
File "C:\Users\nhung\AppData\Local\Programs\Python\Python310\lib\site-packages\torch\cuda\__init__.py", line 314, in _lazy_init
torch._C._cuda_init()
RuntimeError: Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx
Deactivating virtual environment...
'deactivate' is not recognized as an internal or external command,
operable program or batch file.```i used git last time
should i use git or cmd
ill use cmd this time
ornate elk
```C:\AI\ComfyUI>.\start-comfyui.bat
Activating virtual environment...
Running m...
how do your environment path settings look ?
random night
how do your environment path settings look ?
this?
random night
ornate elk
move zluda to a different place like the guide suggested, and edit the path too
nothing should be linked or placed in downloads, that causes permission errors
random night
done
ornate elk
is this bad <@161910921269084160>
the "failed to create model quickly" message can be ignored
its a bug of auto1111 amd
ornate elk
pls full cmd log
ornate elk
its torch load that doesnt work it seems
thats not the ufll cmd log
quick fiber
ornate elk
its using the wrong python it seems
quick fiber
hmm
ornate elk
microsoft store python causes only issues
quick fiber
this is the correct one right
ornate elk
you can also try uninstall every python you have and reinstall only the 3.10.11 64bit with the installer
quick fiber
these are the only two i got
ornate elk
ah okay then we fix it manually
open up a cmd and type
where python
screenshot that
ornate elk
okay
okay launch the python 3.10.11 64bit installer again
click modify, click next, select "add python so environment variables"
then click next
quick fiber
done
ornate elk
okay now open the cmd up again and type
where python
quick fiber
quick fiber
aight
ornate elk
let me know if the webui opens in browser
quick fiber
let me know if the webui opens in browser
will do
my download speeds like 18MB so itll take like 5 mins ish
its been on this for quite a while..
ornate elk
its been on this for quite a while..
ok, let it load
quick fiber
ok, let it load
its done
ornate elk
whats your gpu again?
quick fiber
rx 580
ornate elk
and how much ram do you have?
quick fiber
16
ornate elk
is your gpu driver updated?
also can you try download this model:
https://civitai.com/models/4384/dreamshaper
place it in the models/stable-diffusion folder
delete every other model in there.
Then relaunch
quick fiber
also can you try download this model:
https://civitai.com/models/4384/dreamshape...
i shall try
quick fiber
is your gpu driver updated?
yeah
turbid willow
Uh-oh, I updated my drivers, and it actually took longer to generate an image now XD
When I added one of the lora's before I updated my drivers it took 7 or so minutes to generate an image, and now after updating my driver, the same prompt with the same lora takes 14 minutes.
ornate elk
When I added one of the lora's *before* I updated my drivers it took 7 or so min...
thats not normal by only updating the driver xD
ornate elk
i do have the right driver
okay maybe something is just wrong with directml then
quick fiber
okay maybe something is just wrong with directml then
what shall i do then..?
try the other one?
ornate elk
try the other one?
yep try the zluda version
quick fiber
okay
quick fiber
yep try the zluda version
do i put any other things in args?
ornate elk
quick fiber
ok
@ornate elk where do i find the gfx.zip/ where do i download it
is that the rocmlibs?
ornate elk
is that the rocmlibs?
Yes
quick fiber
im on the last part
hopefully i did everything correct
uh i think i forgot to run webui-user
should i run it and then redo this step?
ornate elk
should i run it and then redo this step?
Yes
You have to do it after the webui-user.bat installed everything
So delete the venv folder and then rerun the webui-user.bat
quick fiber
ok
ornate elk
ok
I also added this step right now so dont be confused if it compiles or crashes while compiling:
- Launch the webui-user.bat
The Compiling of zluda will take 15-40 Minutes depending on the vram and RAM.
The first image will also need to Compile 15-30 minutes before it generates.
After that it doesn't need to compile anymore.
RX580 Users need to relaunch the webui-user.bat everytime the cmd crashes while compiling.
After 15-20 relaunchs the webui should open up and work.
ornate elk
No problem. Let me know how it goes
ornate elk
Nope it would show
Compiling...
ornate elk
You need replace the Library files
ornate elk
Nope
quick fiber
or is this one supposed to be called library and the other old_library
undone robin
Today's API service is not stable, always prompting timeout
undone robin
It seems that there is a serious malfunction. I don't know if it is being resolved.
ruby yoke
It seems that there is a serious malfunction. I don't know if it is being resolv...
There have been cloudflare issues all day
random night
@ornate elkyou're right, everything works just fine when i switch the webui to C drive
ornate elk
<@161910921269084160>you're right, everything works just fine when i switch the ...
Perfect
turbid willow
Should I update my python? I'm using version 3.10.6 and I see that it's super old, like almost 3 years old. I installed the newest version, then I was talking to chat gpt and it told me that 3.10.6 version was actually the best for stable diffusion and that anything past that might break things @_@... But, when using the check version prompt in cmd, it says that I'm still using 3.10.6 despite updating. I told GPT and it told me several things I had to do to change path priorities and stuff and I did those, but it didn't work. Idk if I should keep trying to force it to upgrade or not.
vocal burrow
Should I update my python? I'm using version 3.10.6 and I see that it's super ol...
Only 3.10 ( 3.10.6 and superior ) is guaranteed to be compatible with SD
turbid willow
Sounds good, guess I wont keep trying to upgrade then. Thanks! ^_^
Another issue I'm having, is I;m trying to speed things up by using commandline arguments, but none of them seem to be activating, maybe I'm doing it wrong?
@echo off
set COMMANDLINE_ARGS=--xformers --medvram --cuda-malloc
call webui.bat
if exist webui.settings.bat (
call webui.settings.bat
)
call webui.bat
This is what I have in my webui.bat file, is that not the correct way to add xformers, medvram, and cuda-malloc? I heard those all help so I put them in there.
Doesn't seem to be popping up in the log
Also, it still says that "Hint: Your device supports cuda malloc for potential speed improvements" thing
vocal burrow
@echo off
set COMMANDLINE_ARGS=--xformers --medvram --cuda-malloc
call webui.b...
chatgpt told you to do that ?
you re setting a variable (COMMANDLINE_ARGS) but not using.
turbid willow
Yes, GPT told me to, and also a ton of posts in reddit, and also the stability matrix log recommends it, at least the cuda malloc part.
turbid willow
lol
vocal burrow
Please don t trust it at face value
You re supposed to add command line args in webui-user.bat. There s a line specifically for it
turbid willow
Yeah, it helps "sometimes" but that's also why I search reddit and ask questions here as wlel.
vocal burrow
and you ll see that COMMANDLINE_ARGS is used at the very bottom as an argument for webui.bat
turbid willow
@echo off
set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --xformers --medvram --cuda-malloc
@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
This is what mine looks like, I stuck that xformers and stuff in there too, I put them in 3 differnt files cus I wasn't sure which one was correct lol
vocal burrow
Yeah, it helps "sometimes" but that's also why I search reddit and ask questions...
ok might be used directly in webui.bat then, don t remember exatly haven t opened mine in a while.
vocal burrow
@echo off
set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --xformers -...
you are running this script then and not webui.bat ?
turbid willow
lol, I have no idea what I'm running or doing lol
I just click launch on forge web ui
vocal burrow
I just click launch on forge web ui
what are you clicking exactly ?
turbid willow
vocal burrow
yeah no idea what stability matrix is doing under the hood....
I m not touching that sorry, it just another layer of complexity on top of the whole thing.
turbid willow
Yeah, there's so much complexity, I've learned that over the last week on my noob journey into SD
vocal burrow
I think you can edit the commandline args directly in stability matrix (maybe if you click the three vertical dots or the gear icon)
turbid willow
Tried that, no option in there, unless it's something you have to manually enter. They have a script line or something you can type stuff into.
vocal burrow
Yeah, there's so much complexity, I've learned that over the last week on my noo...
I d suggest you ditch stability matrix then and install things directly.
check the pinned message for some noob friendly guide
turbid willow
I went with stability matrix cus I know nothing about coding.
turbid willow
Maybe, still probably a lot more than my poor brain can handle.
It's very smooth.
Right now I've got things working pretty good, I just have one major issue I'm trying to solve, and that's the problem of Lora's. I can generate a picture in about 20 sseconds, but as soon as I add a single lora, the ETA to generate an image shoots up to 5+ minutes. That's why I'm trying to optimize things right now with the xformers and all the other stuff.
vocal burrow
i can speedrun the install process in probably 30s. you ve got an nvidia card so you ve got it easy.
Download from github,
edit webui-user.bat
run it.
Done
turbid willow
You do make it sound pretty easy. ^_^
turbid willow
Thanks! Yeah, I thought I did it right, they just still don't work though. Shrugs
vocal burrow
screenshot the generation page (with the lora you re trying to use) you might be using things wrong.
(it goes without saying but keep it sfw)
turbid willow
Yeah, I have it added in the webui.bat and stuff.
Oh, well I'm running an illustrious model and I went on the civiai website or whatever it's called and I narrowed search down to illustrious loras, downloaded several, and put them in my lora folder. Then when I have forge open, I see the lora's in the list on the lora page, I click one, it adds it to the prompt, and then my ETA to generate goes through the roof. I've tried several, they all do it.
Very safe for work, don't worry XD
vocal burrow
can you give the link for that model and lora ?
vocal burrow
and what does the cmd log says after a generation?
turbid willow
Is this it?
Also, I don't know why my vram would be low, I have a 4070 12 gig gpu
vocal burrow
the whole log please
turbid willow
Oh, the whole thing? Oof, sec. Feel bad spamming this channel. Sec
Oh wait, I can't even do that. It wont fit in discord. Unless you want me to post like 10x
vocal burrow
That s what this channel is for. I prefer seeing people talking rather than spamming non existing bot for free gen
turbid willow
Omgosh, I think it's like 100 pages long lol
vocal burrow
Oh wait, I can't even do that. It wont fit in discord. Unless you want me to pos...
put it in a .txt and post it there
turbid willow
okay! That's smart
vocal burrow
i ll brb, got to go for a couple of mins
turbid willow
Gpt tells me to fix my vram issues I need to do xformers, medvram, and cuda-malloc. But sadly none of those are working even though I entered all those command lines into my bat file.
vocal burrow
no need for xformers and medvram with forge
turbid willow
Okay.
Well shoot, cus that's what I keep reading might fix my slow generation speeds with loras.
vocal burrow
those are for A1111
turbid willow
I see that A1 steak sauce name gettin' tossed around a lot, but I don't know what it is.
I thought maybe A1111 was part of forge, but I'm just a smooth brained noob XD
vocal burrow
Automatic1111's Stable Diffusion Webui : https://github.com/AUTOMATIC1111/stable-diffusion-webui
aka A1111
it s what forge is based on.
turbid willow
Okay, I see I see.
Sounds like I'm basically hosed and there's no way to fix my problem then I guess.
vocal burrow
I m not used to forge (especially through stability matrix) but I find it weird that it s telling you 4gb of free memory.
vocal burrow
Reboot your computer, make sure that you don t have any game or whatever running in the background eating your vram and try again
turbid willow
I rebooted like 30 minute ago, I have nothing but chrome and discord running.
Even lower now, for memory used
vocal burrow
worse thing you can try to lower your gpu weight value to something like 8gb
should still be enough
turbid willow
Tried, didn't seem to change the eta on image generation
Log is currently saying the same thing.
vocal burrow
you didn t change it mid generation ?
turbid willow
Nah, I interupted the last generation and then lwoered the weight and clicked generate again
vocal burrow
can I see some log with 8gb weight ?
turbid willow
95%|█████████▌| 19/20 [01:00<00:02, 2.99s/it]<01:02, 2.99s/it]
[Low GPU VRAM Warning] Your current GPU free memory is 557.02 MB for this diffusion iteration.
[Low GPU VRAM Warning] This number is lower than the safe value of 1536.00 MB.
[Low GPU VRAM Warning] If you continue, you may cause NVIDIA GPU performance degradation for this diffusion process, and the speed may be extremely slow (about 10x slower).
[Low GPU VRAM Warning] To solve the problem, you can set the 'GPU Weights' (on the top of page) to a lower value.
[Low GPU VRAM Warning] If you cannot find 'GPU Weights', you can click the 'all' option in the 'UI' area on the left-top corner of the webpage.
[Low GPU VRAM Warning] If you want to take the risk of NVIDIA GPU fallback and test the 10x slower speed, you can (but are highly not recommended to) add '--disable-gpu-warning' to CMD flags to remove this warning.
100%|██████████| 20/20 [01:03<00:00, 3.18s/it]
[Unload] Trying to free 4308.86 MB for cuda:0 with 0 models keep loaded ... Current free memory is 363.26 MB ... Unload model KModel Done.
All loaded to GPU.
Moving model(s) has taken 0.02 seconds
[Unload] Trying to free 4257.00 MB for cuda:0 with 0 models keep loaded ... Current free memory is 963.31 MB ... Unload model IntegratedAutoencoderKL Done.
Cleanup minimal inference memory.
tiled upscale: 56%|█████▌ | 5/9 [00:25<00:20, 5.18s/it]
Does my video card not want to give up it's vram to be used or something? lol
vocal burrow
whole log please
turbid willow
Oh, oops, I closed it, the log is gone. Sec, I can make a new one
vocal burrow
looking for some line like that
[GPU Setting] You will use 91.66% GPU memory (11257.00 MB) to load weights, and use 8.34% GPU memory (1024.00 MB) to do matrix computation.
but with 8gb instead of 11
turbid willow
vocal burrow
yeaaaah I don t know.... but when in doubt I blame stability matrix XD
I d still encourage you to check the pinned guide and give the direct forge install a try
you have enough brain juice to check out vram usage on your own so you re already better than half people asking for help here. You should be fine.
turbid willow
Ha, I might have to do that some day when I have some more time to devote to figuring it out.
Thanks for all the help, really appreciate it! ❤️
vocal burrow
Maybe @ornate elk knows something I don t about forge
(incorrect vram usage reported)
turbid willow
Yeah, he's been really helpful.
ornate elk
i heard a few people complaining lately that forge uses more vram than before with the same settings (making it slower), i cant test that myself because the AMD repo is a bit behind.
turbid willow
ornate elk
but i recommended ReForge to the guy complaining and he said it was faster again
turbid willow
Still trying to generate that cat image, using all my vram to do it
vocal burrow
they have one sdxl model 7gb + 1 lora and it s eating up the whole 12gb somehow
not sure what s up with that ....
ornate elk
can you try to remove
--pin-shared-memory
from the arguments ?
vocal burrow
batch size and count to 1 too (re explaining situation for cs1o)
nothing running in the background
turbid willow
I don't think I can remove pin share, cus the bat file ignores things i type in it
ornate elk
they have one sdxl model 7gb + 1 lora and it s eating up the whole 12gb somehow
with hires fix enabled
turbid willow
set COMMANDLINE_ARGS=--xformers --medvram --cuda-malloc
This is what mine looks like, there's not a pin share in there. That's what ur talking about, right?
vocal burrow
if you want to keep using stability matrix you should nt edit .bat manually and use their UI instead.
Otherwise both your modifications and stability matrix will fight each other resulting in who know what depending of order of execution.
turbid willow
That makes sense, problem was I couldn't find anything in the UI to alter those settings so I didn't know how else to do it
turbid willow
Mine doesn't have medvram in there
But, I did find the pin thing in there, I turned it off. I'll try again
vocal burrow
yeah medvram lowvram are A1111 arguments
turbid willow
Hey, I found the cuda thing in there too!
vocal burrow
can t find a screenshot with Launch options for forge
turbid willow
I'll try these few changes, maybe it'll help.
vocal burrow
Hey, I found the cuda thing in there too!
out of curiousity, can you screenshot it plz ?
turbid willow
Uh, sec, sure. It's longer than will fit in a single screen shot though
vocal burrow
just part of it
just want to make sure that it s the proper thing
had too many issues with matrix in the past.
turbid willow
vocal burrow
yeah that looks good
turbid willow
I just unchecked the pin share and clicked the cuda malloc
vocal burrow
edit "command line args" that way going forward
and remove whatever you did in .bat
turbid willow
Yeah yeah, smart, I just did that too.
Lora is still kickin' my poor little cat art images ass...
Your current GPU free memory is 1025.46 MB for this diffusion iteration.
It's higher than it was before though?
I think it was saying like, 500 something MB before?
Though the ETA on the image generation is several minutes longer than before >_>
vocal burrow
can I get one more log now that you ve changerd the args
turbid willow
Sure, sec.
Here is the cat picture to go with the logs XD
I asked for a giant cat in a taco, I didn't get a taco but I wasn't disappointed.
vocal burrow
you can try enabling --always-normal-vram just to see
but with 12gb you shouldn t have to do that...
turbid willow
HOLY CRAP... I tried to generate a second image using the same prompt... I;m getting some cursed image of a cat man with a giant C#&$, wtf? Lol.
Yeah, I can try that, sec, let me bleach my eyes and restart forge.
Hmm, I don't have an always normal vram, I have an always offload from vram though?
turbid willow
Silly Stability. How dare.
When it gives me all these warnings, is it saying there's a risk of harming my GPU, like, am I damaging my GPU here?
[Low GPU VRAM Warning] If you want to take the risk of NVIDIA GPU fallback and test the 10x slower speed, you can (but are highly not recommended to) add '--disable-gpu-warning' to CMD flags to remove this warning.
When they use words like, "risk" and "Fallback" and "highly not recommended" it sounds really bad, yeah?
I fixed it!
I clicked the all button and now it goes fast!
Generated an image in seconds instead of minutes!~
It still says warning, low Vram, but it takes like, 50 seconds to make one image instead of 5-10 minutes
vocal burrow
isn t illustrous sdxl based ?
turbid willow
yeah
vocal burrow
....
turbid willow
I;m generating images sooo much faster now, it's nuts
vocal burrow
i m not sure i get it but ok....
turbid willow
With no lora, I can make an image in like, 3 seconds instead of the 10 or so it took before.
Hmm, and now it slowed down again. Idk, it hates me XD
vocal burrow
what does the UI look like with "all" ? and the log ?
turbid willow
Gpt says that if I switch to async, it will go faster, about to try that after this image is done.
turbid willow
No lol?
vocal burrow
gpu weight at 12 gb
turbid willow
Wantme to lower it?
vocal burrow
yeah let s try 8gb for now. should be enough
ok no
10
let s stick to 10
just to be safe
turbid willow
Sure, when this image is done I'll try with 8gb. I have a stop watching going so I can time speeds.
10gb
vocal burrow
curious also to get log for both 10g and 10g without the lora
turbid willow
It's going faster so far
Oh gosh, I don't know if I know where one log ended and the other started. I'll try
ornate elk
its a much smaller resolution now
vocal burrow
probably too small for sdxl too btw
turbid willow
what is minimum size recommended?
vocal burrow
sd1.5 is trained for 512x512, sd2 for 768x768, sdxl/sd3 1024x1024
turbid willow
If I do less than 1024 will it go slower?
vocal burrow
you can safely prompt for their trained resolution +- 25% pretty safely
turbid willow
Good to know!
vocal burrow
eg, asking for 512x768 with sd1.5 is ok.
turbid willow
But don't go lower than like, 750 for sdxl?
vocal burrow
yes
turbid willow
Gotcha
vocal burrow
if it s too low it won t be able to apply what it learned
turbid willow
Technology is such a trip XD
vocal burrow
if you go too high the output will look "tiled"
like it generated two images in one
turbid willow
I think I've seen that before, yeah.
vocal burrow
so you can get one girl with two torso for instance
vocal burrow
hires.fix allow you to generate at much higher resolution.
tldr it will generate one time at "low resolution", then it will do a second pass using the first result to restrict its output.
vocal burrow
also the denoise strength of hires.fix will pretty much tell "how much to constrain" (the lower the more constrained) that second pass.
0.5 is probably a bit high
0.2/0.3 is probably better.
turbid willow
OKay, I can try to lower that.
But yeah, so far raising the image size looks to have caused way more harm than good
turbid willow
Ooh, I see why it went faster when I switched to "ALL", it's cus it changed all the settings, the sampler, the image size, etc etc
No wonder
I didn't notice it changed all that stuff at first.
idle mica
@ornate elk can you help me and sorry for ping you
turbid willow
I think I got it fixed. If I set this, my lora's don't hose me anymore and I can generate images at a normal 20 seconds or so again.
Idk what this is doing to my quality, but the images look fine so far.
idle mica
i cant rund addeatailer idk why
random night
i cant rund addeatailer idk why
does your cmd has shown any error
random night
i heard a few people complaining lately that forge uses more vram than before wi...
yeah, some people recommend to revert forge to previous version
quick fiber
@ornate elk how do i use a model?
vocal burrow
you put it in the models\Stable-diffusion folder
ornate elk
i cant rund addeatailer idk why
Can you show an example and your cmd log?
quick fiber
you put it in the `models\Stable-diffusion` folder
then where do i choose which one i wanan use
quick fiber
At the top left
does the first compilement take a while or is it always slow
ornate elk
does the first compilement take a while or is it always slow
The first compilation will take a while. But its one time only
quick fiber
The first compilation will take a while. But its one time only
alr
its taking its sweet time
random night
does sdnext don't have those commandline arg such as --ckpt-dir or some sort?
idle mica
Can you show an example and your cmd log?
ornate elk
does sdnext don't have those commandline arg such as --ckpt-dir or some sort?
Yep it doesn't but maybe in the settings
quick fiber
AttributeError: 'NoneType' object has no attribute 'sd_checkpoint_info'
'NoneType' object has no attribute 'sd_checkpoint_info'
Traceback (most recent call last):
File "C:\Users\Rayan\SD-Zluda\stable-diffusion-webui-amdgpu-forge\modules_forge\main_thread.py", line 30, in work
self.result = self.func(*self.args, **self.kwargs)
File "C:\Users\Rayan\SD-Zluda\stable-diffusion-webui-amdgpu-forge\modules\img2img.py", line 248, in img2img_function
processed = process_images(p)
File "C:\Users\Rayan\SD-Zluda\stable-diffusion-webui-amdgpu-forge\modules\processing.py", line 843, in process_images
res = process_images_inner(p)
File "C:\Users\Rayan\SD-Zluda\stable-diffusion-webui-amdgpu-forge\modules\processing.py", line 880, in process_images_inner
p.sd_model_name = shared.sd_model.sd_checkpoint_info.name_for_extra
AttributeError: 'NoneType' object has no attribute 'sd_checkpoint_info'
'NoneType' object has no attribute 'sd_checkpoint_info'
what does this mean
ornate elk
What's the model you tried to use?
ornate elk
There are but normal models can work too
ornate elk
Does txt2img work?
ornate elk
Can you screenshot the settings your used?
quick fiber
idle mica
here cs1
quick fiber
Oh god I just black screened and now my gpu is going vrooooooom
it killed my drivers
my gpu no exist
turbid willow
Woo, did you just nuke ur computer? Damn..
quick fiber
Woo, did you just nuke ur computer? Damn..
idk sometimes if my gpu goes zooooom then it just unistalls itself
hasnt happened in a while tbf
this whole time i thought i had an 8gb gpu
it was 4gb...
quick fiber
yeah i didnt know
quick fiber
Your on a 4gb
is it possible to use Lora models on here?
ornate elk
Could work yes
quick fiber
how do i make it appear on checkpoints?
ornate elk
how do i make it appear on checkpoints?
It goes into models/lora
And then click on the lora tab
There select the lora
quick fiber
And then click on the lora tab
where the tab?
quick fiber
this?
idle mica
Try update your webui and extensions.
Then try with an other model
oki thx
what was the arguments to put on start to it update all the time i open the ai
ornate elk
what was the arguments to put on start to it update all the time i open the ai
Git pull
idle mica
thx
ornate elk
like this right?
Nope.
It has to be in an empty row
ornate elk
<@161910921269084160> then do i put <lora:... whatveer it is?
You click on the lora and it gets added to the prompt
ornate elk
maybe it cant handle this
At which resolution?
Make sure you close all other programms, except discord.
Stuff like games, Wallpaper engine or multiple Browser tabs use vram
quick fiber
512x512
quick fiber
At which resolution?
my gpu keeps crashing causing me to have to reinstall the drivers :/
ornate elk
my gpu keeps crashing causing me to have to reinstall the drivers :/
It shouldn't uninstall the drivers
ornate elk
Yep then press the shortcut
quick fiber
i have tried
ornate elk
It reloads the driver
quick fiber
nothign works
quick fiber
img to img
quick fiber
as soon as i press generate it dies
ornate elk
Would like to check the img2img settings you use
quick fiber
should be all default
zenith loom
guys... need your help on this.. i try to install SD with openvino accelerator since my laptop is kind a potatoes, i do all the instruction in github but eventually the 'accelerated by openvino' script isnt show up at all and the error look like this..any advise on the error ?
quick fiber
@ornate elk
i cant copy and paste the cmd log or it will close
this time it went bblack and then back
with these
ornate elk
with these
Looks normal
ornate elk
Yea probably. Its good that txt2img works xD
ornate elk
I mean you can generate images xD but yea would be much better if you would have more vram
quick fiber
I mean you can generate images xD but yea would be much better if you would have...
yeah 4gb is so little
no wonder i lag on half the games i play
..
i appreciate the help u gave me tho
hopefully i didnt waste too much of ur time
ornate elk
No worries ^^
dense tapir
Hey is there anyway to set up some workflow in Forge where after an image is done generating, it gets sent to IMG2IMG and then it upscales it?
ornate elk
it flashbanged me with light mode lmao
oh xD then set your windows theme to dark 😛
it will apply to firefox too
noble idol
oh xD then set your windows theme to dark 😛
you can do that 😭
sec
better ong
last time i got blinded LOL
sweet thanks 
ornate elk
no problem ^^
undone robin
When will the service be restored?
vocal burrow
That's a question for stability's support service. This channel is community driven, so we can't answer you.
charred rover
fixed
ornate elk
<@161910921269084160> i switched to firefox finally
Dont forget to setup ublock origin too.
The only trustworthy adblocker extension out there.
river yacht
Having some issues getting my stable diffusion to work, if someone has the time to help out for a bit would be an immense help. Currently having issues trying to install the WebUI but I think I might just be going about it all wrong. So question 1 is: Is A1111 even worth installing now? Almost finished the process and then found a lot of people saying it's dead and recommended Forge and ComfyUI instead, opinions?
lucid yoke
Having some issues getting my stable diffusion to work, if someone has the time ...
i recommend forge for a beginner yeah
river yacht
i recommend forge for a beginner yeah
Thank you. Is the process similar to getting a1111 running?
Cause was almost there then it started tossing pything errors about torch around
river yacht
Thank you 🙂
blissful violet
any word on a release of stable for the 50 series gpu? I did the pytorch patch and i can generate images but my tools like Adetailer and controlnet dont work
lucid yoke
any word on a release of stable for the 50 series gpu? I did the pytorch patch a...
hm? patch? i just modified some files in the comfy enviorment
noble idol
Dont forget to setup ublock origin too.
The only trustworthy adblocker extension...
Ah bet I didn't do that
blissful violet
i use SD forge, and i needed to update pytorch and update from cu121 to cu128
final vector
Has anyone else had this issue with SPAR3D? First few lines gives the system specifications.
lucid yoke
Ah bet I didn't do that
how? i told you 😭
final vector
The main error comes from:
NotImplementedError: Could not run 'texture_baker_cpp::rasterize' with arguments from the 'CUDA' backend. This could be because the operator doesn't exist for this backend, or was omitted during the selective/custom build process (if using custom build). If you are a Facebook employee using PyTorch on mobile, please visit https://fburl.com/ptmfixes for possible resolutions. 'texture_baker_cpp::rasterize' is only available for these backends: [CPU, Meta, BackendSelect, Python, FuncTorchDynamicLayerBackMode, Functionalize, Named, Conjugate, Negative, ZeroTensor, ADInplaceOrView, AutogradOther, AutogradCPU, AutogradCUDA, AutogradXLA, AutogradMPS, AutogradXPU, AutogradHPU, AutogradLazy, AutogradMTIA, AutogradMeta, Tracer, AutocastCPU, AutocastXPU, AutocastMPS, AutocastCUDA, FuncTorchBatched, BatchedNestedTensor, FuncTorchVmapMode, Batched, VmapMode, FuncTorchGradWrapper, PythonTLSSnapshot, FuncTorchDynamicLayerFrontMode, PreDispatch, PythonDispatcher].
But I don't know how to resolve this.
lucid yoke
fogor
lucid yoke
o ill not clog up tech support
noble idol
Dms
split jay
hello peeps! maybe i could find some help in here!
i'm having some issues getting the best results with SD+ControlNet.
i did a sketch as a base drawing to make SD and CN paint over it, proposing me different art style eventually... but i can't find out why SD doesn't "inpaint" my drawing. instead, SD is reinterpreting the whole thing.
any ideas why ??
ornate elk
hello peeps! maybe i could find some help in here!
i'm having some issues gettin...
Hey, show your txt2img + controlnet settings
ornate elk
you have selectetd my prompt is more important
change that to balanced and set the weight to 1
if it doesnt work you have to use controlnet-union xl
chilly trail
So I have been trying to use regional promptert in the ForgeUI. all went well and smoothly, until I added two characters loras. The problem is all my gens stops at 95%. with atleast 6-7GB of free VRAM left to be used. I have tried restarting but the same thing happens again. I'm running an rtx 408016GB.
ornate elk
So I have been trying to use regional promptert in the ForgeUI. all went well an...
what does the cmd shows when it stops?
chilly trail
This is at the end: [Memory Management] Current Free GPU Memory: 7581.23 MB
[Memory Management] Required Model Memory: 159.56 MB
[Memory Management] Required Inference Memory: 1024.00 MB
[Memory Management] Estimated Remaining GPU Memory: 6397.67 MB
Moving model(s) has taken 0.07 seconds
Total progress: 100%|| 20/20 [07:57<00:00, 23.90s/it]
Total progress: 100%|| 20/20 [07:57<00:00, 2.65it/s]
ornate elk
This is at the end: [Memory Management] Current Free GPU Memory: 7581.23 MB
[Mem...
okay, make sure the webui is whitelisted in any browser adblocker, also try to enable the Never OOM for "Tiled VAE only" setting at the bottom
chilly trail
As well as this: No target word is detected in Prompt mode
river yacht
Only getting black images after installing my SD with Forge and Flux. Any ideas?
ornate elk
As well as this: No target word is detected in Prompt mode
can you show the complete cmd log?
chilly trail
Will try that. Out of curiosity, why would an adblocker affect my gens? And only when I try to use loras+regional prompter combined?
ornate elk
Will try that. Out of curiosity, why would an adblocker affect my gens? And only...
adblocker also block scripts which some extensions use or the img2img tab of forge and auto
chilly trail
adblocker also block scripts which some extensions use or the img2img tab of for...
Ah. Will make sure to whitelist then. Thanks.
Now it works after enabling OOM tiled vae only setting.
Thank you so much 😄
ornate elk
perfect np 🙂
ornate elk
Only getting black images after installing my SD with Forge and Flux. Any ideas?
which flux model do you try?
river yacht
Uhhh, Nf4 I believe it's called
ornate elk
whats your gpu?
ornate elk
and do you have t5 and clip l and the vae?
river yacht
Explain to me like I am 5
xD
Very new to all this
But I do have a VAE
Running Color101
split jay
you have selectetd my prompt is more important
i just got much better results!
buuuuut it seems that my sketch isnt "really" taken in details. do you reckon controlnet-union xl would solve this problem?
ornate elk
i just got much better results!
buuuuut it seems that my sketch isnt "really" ta...
yea some controlnet models wont work good with xl even if they are made for it
controlnet union was the best so far in terms of compability
split jay
i'll try this! thanks so much for your help
ornate elk
i'll try this! thanks so much for your help
ornate elk
Running Color101
flux needs its own vae
called ae.safetensors
clip t5 and clip l are text encoders needed for flux to work
river yacht
ornate elk
checkout this page for infos:
https://github.com/lllyasviel/stable-diffusion-webui-forge/discussions/1050
ornate elk
https://civitai.com/models/866296/fluxvae this one?
nope, its linked at the page i linked
river yacht
Got it, Will have a look through
river yacht
nope, its linked at the page i linked
Okay following this and cloning the repositories its giving me folders and not files, do I just toss the folders into where they are telling me to?
ornate elk
Okay following this and cloning the repositories its giving me folders and not f...
dont clone repos
they link files you have to download
river yacht
I can't actually find the download to them though
Probably just me having brainfart though
ornate elk
nice, no problem 🙂
river yacht
nice, no problem 🙂
So I got some loras that tell me to edit the "weight" for different things. Where do I find said option?
ornate elk
So I got some loras that tell me to edit the "weight" for different things. Wher...
the lora weight is the number at the end of the lora in the prompt
lora:loraname:1 and 1 would be the weight
river yacht
Gotcha, Thanks 🙂
noble idol
Gotcha, Thanks 🙂
Wait
I want to clarify
Because I've seen people do this
Use the weight on a scale of like 0.1-1
I've seen people do it like
noble idol
lora:loraname:1:1 or soem weird shit
Or use every lora at 1 and wonder why their images aren't coming out right
noble idol
<lora:loraname:1:1> or soem weird shit
Don't do this 
it gave me a stroke last time I saw it
ornate elk
i usually start with 0.8 and see how it goes
blissful violet
running adetailer?
im running a 5070ti and Adetailer won't work for me
it tells me that ADetailer needs to be updated from pytorch 2.6 to 2.7
idle mica
it tells me that ADetailer needs to be updated from pytorch 2.6 to 2.7
ah ok it not only me so thx
i guess we need to wait they update
blissful violet
just gotta wait for that blackwell update
all i need is to update ADetailer and im set
idle mica
all i need is to update ADetailer and im set
sameee
do you know if the creater will update it?
blissful violet
but other than that the generation speed with a 50 series is insane
idle mica
but other than that the generation speed with a 50 series is insane
trueeeee
blissful violet
do you know if the creater will update it?
i did message him and got no response
idle mica
i did message him and got no response
rip
blissful violet
so hopefully im not the only one who got on him
idle mica
hope it update or somone create a new version of it
honest forge
is red bad
blissful violet
yes
blissful violet
is red bad
means there is a error in the prompt
blissful violet
rip
since ADetailer isnt up to date with the 50 series CUDA it'll say the files are corrupted
lucid yoke
since ADetailer isnt up to date with the 50 series CUDA it'll say the files are ...
I use swarm so instead of a detailer i use segment:face and works pretty good. could be a alternative
but you need to mess with a few files in the torch directory to disable torch audio etc
noble idol
i usually start with 0.8 and see how it goes
I do 0.6 to start with usually
Sacrifices so much of thr style though with style loras 😔
ornate elk
Sacrifices so much of thr style though with style loras 😔
yea it depends on the lora for sure
ornate elk
hope it update or somone create a new version of it
@blissful violet for the time beeing you can simulate adetailer by just inpainting the face in img2img.
Mask the face,
set inpaint are to "mask only"
set the denois to 0.5 and press generate
idle mica
since ADetailer isnt up to date with the 50 series CUDA it'll say the files are ...
yappp like me
idle mica
<@138372625658413056> for the time beeing you can simulate adetailer by just inp...
thats a good idea indeed,but dev told anything of update or know that problem?
ornate elk
thats a good idea indeed,but dev told anything of update or know that problem?
you should open up an issue on the adetailer github. The dev was last active yesterday, the repo was last updated 3 months ago
https://github.com/Bing-su/adetailer
ornate elk
Share with the class, I'm curious
i can share some stuff in #🍥｜anime
blissful violet
<@138372625658413056> for the time beeing you can simulate adetailer by just inp...
i tried that and img2img skips it, normally before i got my new gpu i used 5 different detailer tabs for person, face, eyes, hands, and feet if they show
ornate elk
i tried that and img2img skips it, normally before i got my new gpu i used 5 dif...
skips what? inpainting?
blissful violet
it skips the ADetailer
itll do the highlight saying it detected something then nothing
ornate elk
yea because its not updated. i mean you can self inpaint the face in img2img
its in fact all adetailer does
minus the auto face recognition
ornate elk
why in discussions and not in issues 🥲
gray merlin
can anyone help me install comfy on a amd card? i had it before on a nvidia card but i switched and it's more complicated now. i will pay for your time
random night
can anyone help me install comfy on a amd card? i had it before on a nvidia card...
You can just read the pin comment in this section
dense tapir
Hey is there anyway to set up some workflow in Forge where after an image is don...
@ornate elk Is it possible to automate something like this?
gray merlin
OSError: [WinError 126] The specified module could not be found. Error loading "C:\SDZluda\stable-diffusion-webui-amdgpu\venv\lib\site-packages\torch\lib\c10.dll" or one of its dependencies. is my error that i cant get past
random night
OSError: [WinError 126] The specified module could not be found. Error loading "...
do you try delete venv and run webui-user.bat again?
gray merlin
do you try delete venv and run webui-user.bat again?
yes. getting same error
random night
yes. getting same error
could you show full cmd log?
random night
i mean full, from when you start your webui-user.bat to finish
random night
https://learn.microsoft.com/zh-CN/cpp/windows/latest-supported-vc-redist?view=msvc-170 can you try to install this
ornate elk
Open up a cmd as admin and run
Pip cache purge
Then delete the venv folder and relaunch
ornate elk
<@161910921269084160> Is it possible to automate something like this?
Not that I know of
gray merlin
https://learn.microsoft.com/zh-CN/cpp/windows/latest-supported-vc-redist?view=ms...
have it already
gray merlin
Open up a cmd as admin and run
Pip cache purge
Then delete the venv folder and r...
i will try now
ornate elk
Okay now close and relaunch the webui-user.bat
gray merlin
Okay now close and relaunch the webui-user.bat
sweet thanks guys
ornate elk
sweet thanks guys
NP, whats your GPU?
gray merlin
NP, whats your GPU?
Radeon RTX 7900 XT
gray merlin
i tried running comfy but i think i screwed it
ornate elk
Did you Followed my install guide?
gray merlin
i tried im just not very good
gray merlin
everytime i launch it just crashes
ornate elk
Ah okay then something is wrong
Have you replaced the zluda files in the comfyui venv?
gray merlin
i thought i did. i could screen share for you
ornate elk
I'm not at the PC rn
gray merlin
oh okay
ornate elk
You can try delete the venv folder in comfyui and then relaunch
Then you should get an error in CMD
Feel free to share that
random night
yeah it just crashes still
you can run cmd on your comfyui folder, type .\start-comfyui.bat, then you should be able to see the error
gray merlin
im a idiot i dont have nvidia lol
random night
ah yeah, same problems with mine few days ago
gray merlin
my old pc had nvidia amd is so much more confusing
random night
```C:\AI\ComfyUI>.\start-comfyui.bat
Activating virtual environment...
Running m...
here, should help you somehow
scroll down for more info
ornate elk
yep, check your environment path settings, and then replace the zluda files.
im back in maybe an hour if you need help or want to screenshare
supple dawn
How can i set a video for my lock scree on Iphone?
gray merlin
i will be here
gray merlin
am i doing something wrong with biglust? i heard its the best for nsfw but is comes out terrible. i know the prompt is trash but anything you put in comes out awful.
random night
am i doing something wrong with biglust? i heard its the best for nsfw but is co...
what's your cfg scale and denoising strength?
and with sdxl model, i think you need to set your width and height to a minimum of 1024x1024
random night
how about switching to sd1.5 model and see the result yourself?
gray merlin
how about switching to sd1.5 model and see the result yourself?
ill message i dont want to get in trouble for nsfw
ornate elk
am i doing something wrong with biglust? i heard its the best for nsfw but is co...
This server is only for sfw content.
But don't use LCM sampler and check the model version and needed resolution for it
azure magnet
My comfyui can't update to v0.3.26 Can someone help me?
ornate elk
My comfyui can't update to v0.3.26 Can someone help me?
Run a git pull in the comfyui folder or click update comfyui in the comfyui manager
azure magnet
Run a git pull in the comfyui folder or click update comfyui in the comfyui mana...
in C:\ComfyUI or C:\ComfyUI\ComfyUI
And how i run a git pull there?
random night
hmm, can swarm ui use sd1.5 model? cuz i don't see it anywhere?
ornate elk
in C:\ComfyUI or C:\ComfyUI\ComfyUI
in C:\ComfyUI\ComfyUI
click in the folder path then type cmd and hit enter
then type git pull
ornate elk
do you have comfyui manager installed?
azure magnet
do you have comfyui manager installed?
yes
azure magnet
then try update it trough this
I already try update all, and i don't update
ornate elk
I already try update all, and i don't update
what does the cmd shows when you click update
azure magnet
what does the cmd shows when you click update
ornate elk
hm it says comfyui is updated
azure magnet
hm it says comfyui is updated
Could be that i maybe have installed as venv?
ornate elk
Could be that i maybe have installed as venv?
that shouldnt matter
rapid flicker
I tried to install Sygil WebUI manually but got a lot of different errors, each time i fixed one another popped up.
https://github.com/Sygil-Dev/sygil-webui
I just used the Installer version and fixed a few errors like one related to cached_download, I assume one where it needed a token, i manually moved some models over.
There is no webui.bat I was wondering which .py i need to run
anyone who doesnt want dexwallet info? 
lucid yoke
anyone who doesnt want dexwallet info? <:kekwee:1228834195955257424>
dont join its spam
rapid flicker
yeah i figured that out in two seconds
lucid yoke
good lol, sometimes we have people who dont
night salmon
hello i was sent from "MickMumpitz"s Discord to get help here; i always get the issue; SamplerCustomAdvanced MPS backend out of memory (MPS allocated: 9.06 GB, other allocations: 384.00 KB, max allowed: 9.07 GB). Tried to allocate 27.00 MB on private pool. Use PYTORCH_MPS_HIGH_WATERMARK_RATIO=0.0 to disable upper limit for memory allocations (may cause system failure),
im using a low vram workflow and i still get the issue, im so sorry to bother
bleak cloud
Oops! A client-side exception occurred!
Cannot read properties of undefined (reading 'length')
hello tech support what does the below message meant
Oops! A client-side exception occurred!
Cannot read properties of undefined (reading 'length')
thorny lodge
@ornate elk is you DM open?
ornate elk
<@161910921269084160> is you DM open?
depends
thorny lodge
depends
I need help with the stable diffusion, but idk if it's good idea sending here a picture (about AI model)
I got like a monster out of it
ornate elk
I got like a monster out of it
you can send here your settings and the ooutput if its sfw
thorny lodge
you can send here your settings and the ooutput if its sfw
you meant promps?
prompts?
ornate elk
I tried to install Sygil WebUI manually but got a lot of different errors, each ...
its outdated and not supported
thorny lodge
you can send here your settings and the ooutput if its sfw
can also include a pic or rather not?
ornate elk
can also include a pic or rather not?
if its not nudity or gore then yes
thorny lodge
if its not nudity or gore then yes
ammm okay I'm gonna send you my settings:
POSITIVE PROMPTS: (masterpiece, best quality), ultra-realistic, (photorealistic:1.3), 8k UHD, DSLR photo, soft lighting, smooth skin, seductive gaze, natural body proportions, bedroom scene, (perfect face:1.2), NSFW, show boobs
NEGATIVE PROMPTS: deformed, distorted, extra limbs, mutated hands, bad anatomy, blurry, watermark, text, worst quality, low resolution
SAMPLING METHOD: DPM++ 2M SDE
SAMPLING STEPS: 43
CFG SCALE: 7
DENOISING STRENGHT: 0.6 --
I need to make a new model based on the picture i provided
and I got this
my goal is to provide a picture of 1 model and to re-create new model on another location (pool, bathroom, bedroom) and also do NSFW
ornate elk
ammm okay I'm gonna send you my settings:
POSITIVE PROMPTS: (masterpiece, best q...
Set the denois to 0.5 and lower the resolution
thorny lodge
ammm @ornate elk hopefully above is that what you meant?
thorny lodge
Set the denois to 0.5 and lower the resolution
everything else i should leave it?
thorny lodge
Idk your resolution
dimension of the pic should be okay?
thorny lodge
And the model base
can I send that in your DM?
ornate elk
Idk as you didnt showed them
thorny lodge
so I don't get flagged
rapid flicker
its outdated and not supported
recommendation instead? i've been bouncing around changing the versions of various things but it's like error whack a mole. Is there anything with an installer that is a bit more straightforward?
ornate elk
recommendation instead? i've been bouncing around changing the versions of vario...
What's your GPU?
rapid flicker
GTX 1080
ornate elk
GTX 1080
Then you can use Forge or ReForge
Check the pinned messages for the install Guides
rapid flicker
Then you can use Forge or ReForge
That was so much smoother than my hours of going steadily insane playing error whack a mole
do you have a kofi or something?
or is that against the rules 
it's not been solicited so maybe its ok, i want to thank you
ornate elk
do you have a kofi or something?
No problem, glad it worked ^^
And yep I have Kofi its linked at the mainpage of the Guides repo
https://github.com/CS1o/Stable-Diffusion-Info
Also direct linked in my Discord Profile
ornate elk
GTX 1080
With that GPU you can use 1.5 and SDXL models.
But stay away from Flux as its not worth with a GTX
rapid flicker
thank you
ornate elk
No problem 🙂
@rapid flicker I saw that a scammer bot answered you under your initial post.
Dont join their "support server"
Dont give away any personal info
rapid flicker
no i didn't dont worry
turbid willow
Just wanted to say thank you again to all the awesome people here that offer help and support. You helped me so much ❤️ I've been having a blast with stable diffusion now and I couldn't have gotten to this point without you.
night salmon
@ornate elk hello, is your DM open?
random night
what is going on here
merry prism
anyone happen to know how to resolve "'sm_120' is not a recognized processor for this target (ignoring processor)" using nightly pytorch, cuda 12.8, windows10, python 3.12, latest sageattention and triton.. this has been driving me up a wall. tried it in linux as well and was greeted with the same error
idle mica
for the ones that are having problem with adteailer using 50 series this fixed
random night
what is going on here
nevermind, just delete venv file and start over again
ornate elk
<@161910921269084160> hello, is your DM open?
Depends
ornate elk
anyone happen to know how to resolve "'sm_120' is not a recognized processor for...
More context of what you trying to do is needed
ornate elk
for the ones that are having problem with adteailer using 50 series this fixed
Good to know thx. I wonder if it also works if you only add:
set TORCH_FORCE_NO_WEIGHTS_ONLY_LOAD=1
to the webui-user.bat in an empty row at the top.
low sonnet
Hello everyone!
I have an image of character, can you please tell me how i can pose it? I was trying to use controlnet, but the result was bad. Characters look is changing too much. Is it possible to pose with only 1 image of character with controlnet or should I train Lora first to achieve good result, or how can I do it?
vocal burrow
Hello everyone!
I have an image of character, can you please tell me how i can ...
controlnet should be enough, you probably using it wrong if it s not working for you. screenshot your generation settings please.
low sonnet
controlnet should be enough, you probably using it wrong if it s not working for...
send in private
dapper rampart
ohh. ty,also my gpu is 1660ti do you think it will work?
ornate elk
ohh. ty,also my gpu is 1660ti do you think it will work?
Yes with that GPU you should use Forge or ReForge
dapper rampart
i think im downloading automatic1111? im really not sure its for a class
merry prism
More context of what you trying to do is needed
Trying to run hunyuan video on comfyui. I have a feeling either my triton or pytorch instalation isn't properly done considering it won't work with sm120 although everything else is properly installed
hazy wigeon
im looking between 2 graphics cards, asus tuf 3060 12gb vs 3080 turbo 10gb, which one is better for ai rendering?
vocal burrow
3060
in between 3060 and 3080 you ll only get marginal speed improvement. But the 12gb of the 3060 will give you 2gb more to load stuff which is way better. You want to avoid running out of vram at all cost. That s what cause big slow downs.
idle mica
Good to know thx. I wonder if it also works if you only add:
set TORCH_FORCE_NO_...
idk never tryed
ornate elk
i think im downloading automatic1111? im really not sure its for a class
You can use that too.
Checkout the pinned messages for the install guides
hazy wigeon
i think im downloading automatic1111? im really not sure its for a class
what class requires SD lol
just curious
lucid yoke
probably MDI ( Media Design and ICT )
formal ridge
recommendation instead? i've been bouncing around changing the versions of vario...
There's also Lykos Stability Matrix. You just download the exe, run it, tell it what UI you want, then download models from CivitAI (or Huggingface)
lofty cargo
Trying to figure out how the venv mentioned in run.sh comes to be: https://hub.docker.com/r/moophlo/stable-diffusion
Does stable diffusion web ui create one automatically?
for context, I'm trying to adjust this to run comfyUI in a container
candid walrus
OK so long story short I'm getting back into stable diffusion but I also want to game while playing, nothing fancy just like vanilla Minecraft on a server, supper lightweight. Is there a way to limit how much the GPU uses when running the program so the game can have some resources left over for itself? I have a 9070xt anyways and a 7700x with 32gb system ram so I have some compute power to spare.
ornate elk
OK so long story short I'm getting back into stable diffusion but I also want to...
Hey, nope that won't work
You can't limit the vram usage of SD that good
Both, Game and image gen would lag and freeze the PC
Even with a good GPU
vocal burrow
Your best shot would be to force your game to run on the 7700x's igpu. Not sure you want to do that....
ornate elk
It should do it if you use fabric modded Minecraft
merry prism
all out of ideas at this point. using pytorch nightly for cuda 12.8, latest sage attention and triton, cuda 12.8, python 3.12 trying to get hunyuan video running. i've downloaded pre-mades, compiled my own, etc etc. I know i am missing something simple here just can't figure it out. i know it isn't really SD but idk where else to get help with this.
shadow path
Anyone got SD working with a 9070XT yet?
P.S thanks for the AMD guide @ornate elk it's got me this far 🙂
ah hold on, reckon I've fixed it
added these - which one do you reckon fixed it? --no-half-vae --upcast-sampling --no-half
no half?
Guess I'll just let it roll the first generation and try toggling but it's working
yeah, --no-half is the trick
fair oxide
help mee i want to run this thing called Koboldcpp with AMD/Windows/SillyTavern and i dont know which one to pick
craggy orchid
