🤝｜tech-support

[lost dagger]

also should i install this on my local drive or in external drive ?

[ruby yoke]

comfyUI is an interface that you run stable diffusion in. but don't you have photoshop? if you just want to enhance your photos a bit, adobe's gen fill in photoshop is a much better option than stable. it's a lot easier to use and integrated with all the other tools you probably use all the time

[lost dagger]

well i try to do something different with Ai i think that always is a way to make thing different

So you can look diffferent in the market when i say touch up meaning i dont want to make the person look different i could change clothes

i dont know yet because i dont much about this but that just an idea

[ruby yoke]

okay, well - comfyUI is just an interface to run stable diffusion inside of. you can also run stable at the command line, or inside webUI, or auto1111, or forge - or online in a number of places. how much Vram do you have and what GPU do you have?

[lost dagger]

how i know the GPU i have in my computer?

[ruby yoke]

do you have a windows machine, a mac, a linux machine, just your mobile device, something else?

[lost dagger]

mac

[ruby yoke]

okay, you probably want to wait till @ornate elk gets back online and discuss it with him

[lost dagger]

why windows is better for this?

[ruby yoke]

it's not necessarily better, but I don't know how to find out what hardware is in a mac as i don't use macs

[lost dagger]

the GPU for the internet or inside my mac?

[ruby yoke]

gpu = graphics processing unit. so if you're going to run stable diffusion locally, you need one in your machine. if you are going to use something online, then you would be using the GPUS that belong to the site you are using.

but i don't know if macs have different requirements than windows machines, which is why i suggest wait till @ornate elk gets back online

[lost dagger]

Chipset Model: Apple M1 Max
Type: GPU
Bus: Built-In
Total Number of Cores: 24
Vendor: Apple (0x106b)
Metal Support: Metal 3
Displays:
LG HDR 4K:
Resolution: 5120 x 2880 (5K/UHD+ - Ultra High Definition Plus)
UI Looks like: 2560 x 1440 @ 60.00Hz
Main Display: Yes
Mirror: Off
Online: Yes
Rotation: Supported

[fair oxide]

how do you change the output temp folder for gradio? im not a programmer so ull have to say spcific instructions thank you ^^

[quaint finch]

i have some issues using HI red settings, its been fine for months when suddenly when i tried it today, the progress bar is stuck at 0% can anyone help me with this please?

[karmic crown]

how much RAM do you have? our Silicon Macs have a "graphics card inside CPU", which means there is no separate dedicated graphics card as we had on old Macs

they use "unified" RAM, which means it is shared between the CPU, graphics card, and neural engine

key point - sometimes is more important how much RAM you have than which CPU you have
M1 MAX with more RAM might work faster than M3 PRO with less RAM

[calm cairn]

hi again Viking, i thought i would give you an update from earlier:
i found out that, if I took the image I was trying to inpaint and copied it to a new .jpg file, then it would inpaint normally

this implies to me its something in the exif/metadata

anyway thanks for your help earlier, I'm going to try to figure out what settings are wrong on the image now 🙂

[karmic crown]

there is no "better" or "worse", they work the same
the key difference is speed
windows with newer NVidia (4090) works faster than any Mac

[karmic crown]

to install stabile diffusion webui on Mac follow my step by step guide or simply use my a1111-setup script that will do everything for you

yesterday i found out that the default model that was automatically downloaded by webui has been deleted (https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/16457)

i will try to update my guide later today (i need to go now) and release a new installer version with a patch for that issue

you can manually download a model as described in the answer on that link

just be sure to put it in stable-diffusion-webui/models/Stable-diffusion folder

[karmic crown]

CS1o and Aryetis know much more about NVidia and AMD than me, but for Macs, it's always best to wait for me 🙃

[ornate elk]

Can you show the settings you used and the cmd log?

[quaint finch]

the problem was so sudden its quite surprising

months, alls fine

suddenly BAM! it stopped working

[ornate elk]

Ah I see, when using hires fix you have to set the hires steps to 10.
Also don't use a refiner

[quaint finch]

i see

but ive used the same settings for a long time

and it hadnt shown any issues for months

[ornate elk]

What's your GPU?

[quaint finch]

3070

[ornate elk]

Ah okay. Then also make sure you have --xformers --medvram-sdxl --no-half-vae in your webui-user.bat for the best performance

[quaint finch]

ill have that set up, and see how it works

thanks @ornate elk

[ornate elk]

Refiner is not needed at all as hires fix is better.
Refiner was only useful for the base sdxl model.

[quaint finch]

i understand

[ornate elk]

But never leave the hires steps at 0, better use 10-15

[quaint finch]

🙏

[ornate elk]

You used a shortcut here to launch. Make sure you add the commandline_args I mentioned there and in the original webui-user.bat

[upper moth]

Hey everyone! I used to be able to run SD without much issue but I recently upgraded my PC and now I'm having trouble getting it to even launch. I’ve hit a wall with errors but before we get into that, I'd like to ask if it maybe is a compatibility issue with some of my specs:

System Specifications:

• CPU: AMD Ryzen 5 7600X
• GPU: AMD Radeon RX 7800 XT
• RAM: 64GB DDR5
• Storage: 1TB SSD
  Software Environment:
• Operating System: Linux Mint 22

Is any of those a big deal breaker? I'll send more info if needed!

[ornate elk]

Hey, what did you upgrade?

Or did you reinstalled the OS

Can't help much with Linux but checkout AMDs ROCm guide

[upper moth]

the machine is different, like it's a fresh start basically

[ornate elk]

Here:
https://rocm.docs.amd.com/projects/install-on-linux/en/latest/install/quick-start.html

After Installing ROCm you have to git clone the stable-diffusion-webui

[upper moth]

okay, out of the gate it looks like my kernel may be an issue, the website lists 6.8[GA] as the compatible one, but I have 6.10, I'll see if I can do something about that

thanks for the pointer!

[strange zephyr]

can someone tell me how I can use LDSR upscaler here? Ignore the LDSR.ckpt in the list, I downloaded this and placed it in the ESRGAN folder but it's not the correct model. Is there a special version of LDSR I need to download to use it as an upscaler here? In Auto111 I was able to just chose LDSR from a list when using Ultimate SD upscale...

[karmic crown]

I have updated my guide for Macs https://github.com/viking1304/a1111-setup/discussions/2 to address the issue with the download of the default model

i do not have time to update the installer, it works, just the model needs to be downloaded manually, until I apply the patch

[thorn pendant]

some how when i restart pc it del my added --xformers --medvram --no-half-vae
how to fix this

[ornate elk]

Add it to the commandline_args

[thorn pendant]

i addded and save but when i resstart pc it become normal

that's problem

[ornate elk]

that makes no sense

but you can right click the file and enable "Write Protection" after you edit it

[chrome rampart]

hey all hope doing well. Since to i am get lots of error like: Using pytorch cross attention
Using pytorch attention for VAE
ControlNet preprocessor location: F:\a1111-forge\webui\models\ControlNetPreprocessor
*** Error loading script: animatediff.py
Traceback (most recent call last):
File "F:\a1111-forge\webui\modules\scripts.py", line 525, in load_scripts
script_module = script_loading.load_module(scriptfile.path)
File "F:\a1111-forge\webui\modules\script_loading.py", line 13, in load_module
module_spec.loader.exec_module(module)
File "<frozen importlib._bootstrap_external>", line 883, in exec_module
File "<frozen importlib._bootstrap>", line 241, in _call_with_frames_removed
File "F:\a1111-forge\webui\extensions\sd-forge-animatediff\scripts\animatediff.py", line 10, in <module>
from scripts.animatediff_infv2v import AnimateDiffInfV2V
File "F:\a1111-forge\webui\extensions\sd-forge-animatediff\scripts\animatediff_infv2v.py", line 5, in <module>
from ldm_patched.modules.model_management import get_torch_device, soft_empty_cache
ModuleNotFoundError: No module named 'ldm_patched.modules.model_management'

any idea since i dont think i use anything with animation?

[ornate elk]

animatediff works only on the old forge version

[chrome rampart]

hey cs. Do i need to do anything like remove extension or something or just error and let it be ?

[ornate elk]

To not get the error you need to remove the extension

[chrome rampart]

is that extension forge_space_animagine_xl_31 ?

[ornate elk]

No its sd-forge-animatediff

[chrome rampart]

sorry i think found it must be sd-forge-animatediff

[chrome rampart]

super thks help again as always. You guys are awesome

[ornate elk]

no problem 🙂

[inner dawn]

Help me, please, what to write in webui-user SD Forge, so that it can see the ControlNet models that are in A1111.
Now it is as on the screenshot.

[ornate elk]

Hey, you have to edit the webui-user.bat of Forge like this:
`@echo off

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

@REM Uncomment following code to reference an existing A1111 checkout.
set A1111_HOME=D:/Programme/AI-Zeug/stable-diffusion-webui-amdgpu
@REM
@REM set VENV_DIR=%A1111_HOME%/venv
set COMMANDLINE_ARGS=%COMMANDLINE_ARGS% ^
--ckpt-dir %A1111_HOME%/models/Stable-diffusion ^
--hypernetwork-dir %A1111_HOME%/models/hypernetworks ^
--embeddings-dir %A1111_HOME%/embeddings ^
--vae-dir %A1111_HOME%/models/vae ^
--controlnet-dir %A1111_HOME%/models/controlnet ^
--lora-dir %A1111_HOME%/models/Lora

call webui.bat`

[ruby yoke]

Will make a note of that 🙂

[inner dawn]

Thank you. Strange that it now see all controlnet models except this -diffusion_pytorch_model_promax.safetensors
https://huggingface.co/xinsir/controlnet-union-sdxl-1.0/tree/main

[steel idol]

Hi could someone help I have difficulty installing sd I have never use GitHub and its really hard

[sinful marlin]

What about just VRAM sharing?

[ornate elk]

Nope

[ornate elk]

Hey, checkout my install guide on the pinned messages

[steel idol]

Okay thanks😊

And so when I’m here and I have type the prompt and click generate I just need to wait ?

[steel idol]

The cmd wrote “stable diffusion model failed to load”

[ornate elk]

You need to download a model from Civitai.com

Put it into the models/stable-diffusion folder

[steel idol]

Okay thanks

It’s a .safetensors files?

[ornate elk]

Yep

And needs to be at least 2gb

[steel idol]

Now it wrote “invalid argument to memory_allocated” 🥲

[ornate elk]

What's your GPU?

Can you show the cmd log?

[steel idol]

It’s an rx6600

[steel idol]

How ?

[ornate elk]

Copy and paste ^^

[steel idol]

It’s too long for discord😭
Sorry I’m not really smart 😂

venv "C:\Users\mateuhla\stable-diffusion-webui-amdgpu\venv\Scripts\Python.exe"
ROCm Toolkit 5.7 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-5-gd8b7380b
Commit hash: d8b7380b18d044d2ee38695c58bae3a786689cf3
Using ZLUDA in C:\Users\mateuhla\stable-diffusion-webui-amdgpu.zluda
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
C:\Users\mateuhla\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=CPUExecutionProvider, available=['TensorrtExecutionProvider', 'CUDAExecutionProvider', 'CPUExecutionProvider']
ZLUDA device failed to pass basic operation test: index=None, device_name=AMD Radeon RX 6600 [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.

[steel idol]

Is this enough ?

[lost dagger]

@ornate elk can you help me out with my question

i dont know if you swa my questions

[ornate elk]

Did you replaced the rocm lib files ?

Also show your environment path settings

[ornate elk]

What's the issue?

Didn't saw it

[steel idol]

No I have just download it

[ornate elk]

Yea thats the issue then

You have to do step 4-5 of the guide at zluda setup

Replacing some files In a folder

[steel idol]

Oh okay thanks 😊

[lost dagger]

i wanna know if i should install this on local or if any other way not too hight prices that i dont need to have that in my drive ?

or if can install that in external hardrive

[ornate elk]

You can install it localy on your Mac with the easy install script. Its listed in the pinned messages of this channel.
Then you can decide if its to slow to your liking

[ornate elk]

Online services like Leonardo.ai are good too. Or you can rent a GPU online too

But you should try it out localy first

[ornate elk]

Idk. For mac its a bit difficult I think. The install script installs it on the main drive by default

[lost dagger]

ok thank

@ornate elk also if i trying to level up my photography and video work what type of model pack should i loook for

or where can i read and learn more about this so i know what i gonna be doing?

[steel idol]

I think I don’t understand this part well: 3.3 Go into C:\Program Files\AMD\ROCm\6.1\bin\rocblas folder.
There copy and rename the library folder to old_library
Open the .zip file and drag and drop all files of the library folder into the library folder. Not into the old_library folder.
Important Restart your PC after this Step before proceeding.

[ornate elk]

in your case its the 5.7 folder

[steel idol]

And so I copy the library file so there’s two in robclas ?

[ornate elk]

yep

and then you rename one

to old_library

and then you move the files you downloaded into the library folder (not old_library)

[steel idol]

And I replace or ignore the identical files ?

[ornate elk]

replace

[steel idol]

Okay

[ornate elk]

and after that restart the PC

[steel idol]

And after this it is supposed to work ?

[ornate elk]

yep

if everything is correct

[steel idol]

This don’t work 🥲
Sd model failed to load

[ornate elk]

can you copy the whole cmd log?

mark everything and right click to copy

[steel idol]

its should be that i think

yeah it is okay sorry i was on phone

[ornate elk]

something isnt correct still:
ZLUDA device failed to pass basic operation test: index=None, device_name=AMD Radeon RX 6600 [ZLUDA]
try deleting the venv and the .zluda folder that are inside the stable-diffusion-webui-amdgpu folder and then relaunch the webui-user.bat

[steel idol]

Okay looks like it work, thanks

[solemn girder]

Stable diffusion on my end is not generating nsfw rn and idk why.

[ornate elk]

Perfect np

[vast pulsar]

Hey guys is if possible to do this with the image 2 image function?

Basically I wanna change a texture from a yellow like skin to pinkish

[stark dew]

اه

hi

[ruby yoke]

that would be an inpainting function, where you mask out the areas of the skin, and then prompt for a different color skin

[sinful marlin]

For some reason when I use --medvram it'll be just about to complete the image and then Stable Diffusion just craps out

[ornate elk]

whats your gpu?

[sinful marlin]

I also have FP8 Weight Disabled

Tesla P4

8gb vram

[ornate elk]

how much vram?

make sure you use --xformers --medvram-sdxl --no-half-vae

and enable fp8 mode for sdxl in settings

[sinful marlin]

I'll give that a try

[sinful marlin]

Does the order matter?

[ornate elk]

nope

[sinful marlin]

So it's able to generate images without issues but my seeds don't seem to work :/

Should I try disabling the fp8 mode for sdxl?

huh, now whenever I try to disable fp8 mode for sdxl my instance crashes

[ornate elk]

show me the settings when you get this?

[sinful marlin]

And when I start it up again it's back to being enabled

[ornate elk]

try load an 1.5 model then (2gb) and then try disabling fp8

[sinful marlin]

I'll be back soon to test that

[frigid crypt]

Im running forge with flux just installed upon generating the first image i get this error:
AssertionError: You do not have CLIP state dict!

[ornate elk]

which flux model did you tried? and which webui?

[frigid crypt]

https://huggingface.co/black-forest-labs/FLUX.1-dev/tree/main

webui forge

[ornate elk]

then use this model:
https://huggingface.co/lllyasviel/flux1-dev-bnb-nf4/tree/main
the flux nf4 v2 (best for forge)

or the 16gb flux version:
https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors

[frigid crypt]

will the latter work with 10gb RTX?

[ornate elk]

yes but slow

the nf4 should work much better

[frigid crypt]

whats the diference between the one i tried using at first?

[ornate elk]

its the 23gb base model

wouldnt try to run it on 10gb vram xD

[frigid crypt]

godammit

so should i delete it completely?
for disk space?

[ornate elk]

yep

you dont need it

@frigid crypt also here is a guide on how to use different flux models with Forge:
https://github.com/lllyasviel/stable-diffusion-webui-forge/discussions/1050
you also need to download clip l and t5 for flux

[last tide]

Im trying to test out the flux.1-dev aswell as SwarmUI but i need a flux vae, where can i find it and where do i place it?
Downloaded both file in this image and placed them in models/VAE but doesnt work like that apparently.

[frigid crypt]

whats the he flux1-dev-fp8 one?

[ornate elk]

here is the vae:
https://huggingface.co/black-forest-labs/FLUX.1-schnell/blob/main/ae.safetensors

drop it into models/vae

remove the other 2 files

[last tide]

Will do, thank you

[ornate elk]

?

[frigid crypt]

flux1-dev-fp8 model

[ornate elk]

yea thats the name

its the model i linked [above](#🤝｜tech-support message)

[frigid crypt]

is it good with 10gb vram?

[ornate elk]

its slow but works

better use the nf4 model xD

[steady thistle]

Oh God, my ComfyUI crashed and died yet again.

Trying to load a big Outpainting workflow, it downloaded lots of stuff and then died. ComfyUI will not start up. I'll start with the venv and get rid of some nodes or something.

[frigid crypt]

which clip-l and t5-xxl should i download? or all of them?

[steady thistle]

Maybe.... @ornate elk since I installed the venv before I got my torch installed properly, that I needed to restart the venv and that'll fix everything? Wishful thinking, but I have no idea.

[last tide]

I dont suppose this is normal? Its crashing after maybe 2 minutes of generating an image. using swarmUI with Flux1-dev

[ornate elk]

only the clip l and the t5 fp8

[ornate elk]

which torch version did you installed?

[ornate elk]

whats your gpu?

[last tide]

7900 gre

amd

[steady thistle]

I ran this: pip install -U xformers --index-url https://download.pytorch.org/whl/cu121

[ornate elk]

then you have to do this after the venv gets recreated

[last tide]

at the beginning of the install i selected amd version

[ornate elk]

that installed only the directml version of swarm and comfyui, not the zluda version.
which means it cant run flux as it takes to much vram

[steady thistle]

Okay, I'll let the venv finish recreating then I'll run that command again?

[frigid crypt]

still following the guide:
do I need GGUF models?

[ornate elk]

nope

if you have the nf4 v2 and the flux dev fp8 16gb your fine

[frigid crypt]

whats the best res for the model?
the default res currently takes 55 secs

[last tide]

i still have the zluda files from the stab-diff, any idea how i can get it to work?

[ornate elk]

thats already a good resolution. the higher you go the longer it will take

[ornate elk]

i currently dont have a guide for swarm, but i have a guide for comfyui and forge which both can use flux

with zluda

[last tide]

Guess ill just have to install comfyui then and learn it xD. WIll install it tomorrow

[steady thistle]

I got this error. Then I got a popup saying this: The procedure entry point could not be located in the dynamic link library D:Ai\ComfyUI_windows_portable\ComfyUI\venv\Lib\site-packages\torchvision_C.pyd.

[ornate elk]

got Swarm working with Comfyui Zluda

will update my guide tomorrow. But basicly you need to have Comfyui with Zluda installed first. then you can import it into swarm ui

[shell stag]

Anyone can help me I downloaded extension which is wildcard gallery but when I add wildcard their preview doesnt show up can anyone pls help

[steady thistle]

@ornate elk Perhaps installing this file would help? That file isn't there

Not that I know where to even find that file

[steady thistle]

I looked and that file is in the right folder so the error makes no sense.

Activating virtual environment...
Running main.py with --auto-launch argument...
[START] Security scan
[DONE] Security scan

ComfyUI-Manager: installing dependencies done.

** ComfyUI startup time: 2024-09-14 19:30:09.370774
** Platform: Windows
** Python version: 3.10.11 (tags/v3.10.11:7d4cc5a, Apr 5 2023, 00:38:17) [MSC v.1929 64 bit (AMD64)]
** Python executable: D:\Ai\ComfyUI_windows_portable\ComfyUI\venv\Scripts\python.exe
** ComfyUI Path: D:\Ai\ComfyUI_windows_portable\ComfyUI
** Log path: D:\Ai\ComfyUI_windows_portable\ComfyUI\comfyui.log

Prestartup times for custom nodes:
0.0 seconds: D:\Ai\ComfyUI_windows_portable\ComfyUI\custom_nodes\rgthree-comfy
0.0 seconds: D:\Ai\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Easy-Use
0.7 seconds: D:\Ai\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager

Total VRAM 16376 MB, total RAM 32536 MB
pytorch version: 2.4.0+cu121
D:\Ai\ComfyUI_windows_portable\ComfyUI\venv\lib\site-packages\xformers\ops\fmha\flash.py:211: FutureWarning: torch.library.impl_abstract was renamed to torch.library.register_fake. Please use that instead; we will remove torch.library.impl_abstract in a future version of PyTorch.
@torch.library.impl_abstract("xformers_flash::flash_fwd")
D:\Ai\ComfyUI_windows_portable\ComfyUI\venv\lib\site-packages\xformers\ops\fmha\flash.py:344: FutureWarning: torch.library.impl_abstract was renamed to torch.library.register_fake. Please use that instead; we will remove torch.library.impl_abstract in a future version of PyTorch.
@torch.library.impl_abstract("xformers_flash::flash_bwd")
xformers version: 0.0.27.post2
Set vram state to: NORMAL_VRAM
Device: cuda:0 NVIDIA GeForce RTX 4070 Ti SUPER : cudaMallocAsync

That's when it crashes

@ornate elk Okay, new info. "Torch not compiled with CUDA enabled" ... I tried to start ComfyUI without using your .bat file but with the Comfy run_nvidia_gpu.bat and got that error

[steady thistle]

Oh damn, I may not have CUDA installed. I thought it was automatically installed on the PC

[steady thistle]

Anyone know if I need to install CUDA?

[steady thistle]

Okay, that was the wrong tree to go down. For some reason, there's a mismatch between my torch install and xformers

[unborn bramble]

Can you use the Flux Lora trainer with a T4 GPU? I do have colab pro but I'd rather not use 5 compute units per hour

[unborn bramble]

And then I realized you can't pick a base model for that trainer

Is there a colab that trains Flux loras and allows you to chose your own base model

[sinful marlin]

Yup that fixed it!

[jagged rampart]

What can you tell me about “search-and-recolor” in the “stability ai api”! It can be seen that the colors specified in the prompt are not uniform in the output for each reference image. Based on the above, is it possible that the color of the original image is affecting the output result, is this correct in your perception? In the future, we would like to make detailed color settings using color codes, but is there a setting that allows us to output the specified colors as they are?

[thorny iron]

Would someone help me HD'ify a picture's hair?

[heady pine]

I'm trying to use Flux on Forge UI, but when I try and use DPM++ Karras, it crashes, here is the cmd

[fair oxide]

hi everyone. did someone use 3D photo inpainting node before ?

[shut geyser]

if i have an image which i know is ai generated, can i somehow know what model was used to make it? or in which software?

[ornate elk]

Flux doesn't work with most samplers

[ornate elk]

You can check its metadata. If the creator didnt removed them, there you get the info

[heady pine]

Sick, alright thank you

[last tide]

Ill keep my eyes open

[shut geyser]

if someone removed them, can i still somehow have a clue?

could i send you the image in dms to see if you have an idea?

[opaque star]

Hello, where can i find API organization ID?

cannot find it anywhere in API dashboard

[tranquil marsh]

Hi, does the Flux Realism lora work in SD Forge with flux1-dev-bnb-nf4-v2.safetensors? With the same seed I see no difference in outputs, however I'm not sure if there are trigger words to use alongside just including the Lora in the prompt?

[ornate elk]

No you only can guess, sure you can send it to me

[steady thistle]

I'm using torch 2.4.0+cul21, is this going to be causing me issues with some nodes?

I want to run some outpainting stuff but I'm really getting gun shy about loading nodes because of the crashes. I'd really like to get this working properly.

[ornate elk]

Checkout the nodes on github before installing them. Check for their requirements and how updated they are in general

[steady thistle]

Hmm, perhaps I should do a fresh ComfyUI install. I've been loading workflows so I can understand the steps better, but that means I was loading old nodes sometimes.

[steady thistle]

Okay, new ComfyUI and I see one warning but no errors. Being careful now to only install custom nodes that aren't depreciated or have other issues.

it has this warning: D:\ComfyUI2nd\ComfyUI\venv\lib\site-packages\torch\nested_internal\nested_tensor.py:417: UserWarning: Failed to initialize NumPy: _ARRAY_API not found (Triggered internally at ..\torch\csrc\utils\tensor_numpy.cpp:84.)
values=torch.randn(3, 3, device="meta"),
Using pytorch cross attention
[Prompt Server] web root: D:\ComfyUI2nd\ComfyUI\web

Loading: ComfyUI-Manager (V2.50.3)

ComfyUI Revision: 2695 [e813abbb] | Released on '2024-09-15'

Is that "failed to initialize NumPy" important?

[steady thistle]

So I made a new ComfyUI and used your Start-ComfyUI.bat and it installed torch 2.4.1 at first, then it proceeded to uninstall torch and now in the venv environment it says torch is version 2.3.0. Is this how it should be?

The bat itself says to install old versions of torch and I don't understand why it would do that.

[ornate elk]

you have to edit the bat

to install torch 2.4.0

by default it installs 2.3.0

[steady thistle]

Okay! I will do that. Do I go with 2.4.0 or 2.4.1?

[ornate elk]

I dont know

[steady thistle]

All good! I can test. Much appreciated

Well that loaded super fast and no warnings

[elder night]

Ishsisj

Jssj hi usksjksjsksjsksikw

😁😊

[elder night]

Ienslnssj

[last tide]

Are you a damaged ai model sir?

[fair oxide]

Does anyone know how to setup the Hyper-SDXL-8steps-CFG-lora? I could not get any good results using it.

[steady thistle]

torchaudio and torchvision need torch 2.4.1 but xformers needs torch 2.4.0 . I think this is my big issue. I must not be pulling the most recent version of xformers.

[ornate elk]

Its there now! https://github.com/CS1o/Stable-Diffusion-Info/wiki/Installation-Guides#amd-swarmui-with-zluda

SwarmUI with ZLUDA install Guide

[ornate elk]

then go for 2.4.0

[steady thistle]

I found a newer version of xformers but I don't know how to compile it properly. It's at https://github.com/facebookresearch/xformers/releases

but I'll try to go for 2.4.0 but I swear I tried that earlier. I'll just try again.

My internet was no longer metered as of last week, thankfully. Here goes 2.7gb again.

[steady thistle]

Okay, I think I'm closer. I found I was installing torch 2.4.0 from https://download.pytorch.org/whl/cu118 but xformers from https://download.pytorch.org/whl/cu121 . These should be the same version?

[ornate elk]

yes that needs to be the same

best would be cu121

[steady thistle]

Okay

[last tide]

Even if you do have amd you should click no?

[ornate elk]

Yes

[last tide]

Alright thanks, got a bit confused as of why i should select no :))

[ornate elk]

Yea because when selecting AMD it would install the directml. And No, will install torch which we need for zluda later

[last tide]

Ahhh thats why it didnt work for me earlier aswell

[steady thistle]

Brand new error doing a simple image generation: Numpy is not available.

I did pip list and found numpy 2.1.1

I googled and changed the version and installed numpy==1.26.4

Success

[last tide]

I did copy and rename them like you wrote in the comfyui tutorial aswell.

[ornate elk]

Have you done the comfyui guide before?

[last tide]

I tried it before and it failed. Deleted the folders tho

I could reinstall the comfyui and paste it again see if it will work

Gonna try and reinstall comfyUI

[steady thistle]

bangs his head Using ComfyUI Manager, if I restart through the manager, it rebuilds the entire venv.

Then kills the venv and won't run again until I rebuild everything.

[ornate elk]

What failed?

[last tide]

I saw on github its because i need a specific hip-sdk version

5.7.1

[ornate elk]

What's your GPU again?

[last tide]

7900 gre

[ornate elk]

Then you dont need 5.1

Follow my Comfyui guide not the quick install

[last tide]

Aight, ill delete em and retry

[untold spoke]

hey im trying to replicate this app https://mnml.ai/ can anyone help me how to get it with sd

[hazy canyon]

Hi guys i have nvidia 12Gb ddr5 and i install story diffusion but at the final says torch.OutOfMemoryError: CUDA out of memory. Tried to allocate 864.00 MiB. GPU 0 has a total capacity of 12.00 GiB of which 0 bytes is free. Of the allocated memory 25.46 GiB is allocated by PyTorch, and 812.84 MiB is reserved by P
yTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. and i look all .py to find but i can't... were is located?? thx

[ornate elk]

hey, you need to edit your webui-user.bat
at the line Commandline_Args= you need to add: --xformers --no-half-vae
then save and relaunch

[hazy canyon]

i use a external program to intall sttable difsion --- i use pinokio thx for answer me gonna try to do it im new in this world thx

[ornate elk]

ah okay, no problem

the webui-user.bat is inside the stable-diffusion-webui folder

[hazy canyon]

thx but pinokio just download repo and execute just have the folder storydiffusion-comics.git then in sub-folder : app --- cache and some json files i use pinokio becouse i can have in order my repo i use note ++ and i gonna search to change i make troubles for me so easy... 😆

[ornate elk]

oh okay

i tought you installed stable diffusion

[hazy canyon]

i think that is more easy than have a program that cant change anything thx for all if i can solve gonna put here if someone have my problems thx a lot

[heady pine]

I am having a bit of an issue with generating pictures with Flux,

I have installed the VAE and text encoders like I'm supposed to and put them in the folders where they belong.

I mainly use SD 1.5, and I never have issues with wait times like this,

My GPU is a RTX 2080Ti.

The GPU weights, swap method, and swap location are set to what they are by default, so I don't know if they have anything to do with it.

[stuck flume]

When using SD upscale ESRGAN_4x i get very different time scales on generation. Up to 3,5x goes in less than 3 minutes. If I try with 4x, it takes 4-6 hours to finish. Is this because of some software setting or because of physical memory? I'm running a 4090 24 Gb.

[ornate elk]

make sure you only use the nf4 v2 or the fp8 16gb flux model

all other ones are very slow

[ornate elk]

you need to make sure to set a hires step value

like 10-20

[stuck flume]

Sorry, that flew right over my head. What is a flux model?

[ornate elk]

that message wasnt for you

[stuck flume]

I usually scale with 50 steps using DPM++ 2m SDE Karras (35 for the base generation, higher for upscaling).
I will try lesser steps.

[stuck flume]

Ooops, sorry

[heady pine]

Alright, thanks. I will try the fp8 16gb flux model. I have seen some amazing pictures that people have made with the flux model/s and I would like to at least try my hand at making something. Hopefully this one works. Do you need something similar to a 4090 to be able to use the 22gb Flux model?

[ornate elk]

the nf4 makes also good images. its the same as fp8 just better optimised

[heady pine]

could you give me a link to that one please, I looked and there are a few with the nf4 in the title

[ornate elk]

here you need the v2
https://huggingface.co/lllyasviel/flux1-dev-bnb-nf4/tree/main

[heady pine]

Thank you.

[last tide]

Alright, followed your guide. Got a small different warning/error now. Am generating my first prompt as well so lets see if it works.

[ancient drift]

You can 1 click install SwarmUI on Stability Matrix

If you're having trouble with a manual install

[last tide]

I got amd so need zluda, also gonna follow CS1o's advise cuz been struggeling for few days now with all sorts of things

[ancient drift]

Was just a thought. Only just joined so i dont know the history of your problems. Apparently Stability Matrix only supports AMD on linux, for swarmUI also.

[last tide]

Ah alright and welcome to the club :)

[ancient drift]

Cheers 😄

[last tide]

Hmm ComfyUI execution error: Numpy is not available

guess numpy isnt installed correctly?

downgrading and upgrading numpy does not solve this (tried 1.26.4, 1,24,1 and now 2.1.1)

[ruby yoke]

make sure your path statements are set up correctly

[ruby yoke]

i don't see anything about python, or numpy, in that?

[last tide]

They are, also stable diffusion works fine

[stuck flume]

Nope, didn't work. Even with 10 steps it says 18 hours to generate.

[ornate elk]

Can you generate?

[last tide]

Yes it starts to generate by the looks of it and then comes up after 30 seconds or so with the error

currently on numpy version '1.26.4'

[steady thistle]

I had this error /just/ today. I think there's a change somewhere, but what I did was pip install numpy==1.26.4

Oh, I had numpy 2.1.1

[last tide]

sadly i am already on that version

[steady thistle]

Try install numpy 2.1.1? lol

[last tide]

Already tried that without any success

imma just reinstall numpy completely

Nope, its generating and then BOOM error

[ornate elk]

It crashes with numpy and then no image is generated?

[last tide]

Correct

22:38:00.926 [Warning] [ComfyUI-0/STDERR] Traceback (most recent call last): 22:38:00.927 [Warning] [ComfyUI-0/STDERR] File "H:\SD-Zluda\SwarmUI\dlbackend\comfy\ComfyUI\execution.py", line 323, in execute 22:38:00.927 [Warning] [ComfyUI-0/STDERR] 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) 22:38:00.927 [Warning] [ComfyUI-0/STDERR] File "H:\SD-Zluda\SwarmUI\dlbackend\comfy\ComfyUI\execution.py", line 198, in get_output_data 22:38:00.928 [Warning] [ComfyUI-0/STDERR] 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) 22:38:00.928 [Warning] [ComfyUI-0/STDERR] File "H:\SD-Zluda\SwarmUI\dlbackend\comfy\ComfyUI\execution.py", line 169, in _map_node_over_list 22:38:00.929 [Warning] [ComfyUI-0/STDERR] process_inputs(input_dict, i) 22:38:00.929 [Warning] [ComfyUI-0/STDERR] File "H:\SD-Zluda\SwarmUI\dlbackend\comfy\ComfyUI\execution.py", line 158, in process_inputs 22:38:00.929 [Warning] [ComfyUI-0/STDERR] results.append(getattr(obj, func)(**inputs)) 22:38:00.930 [Warning] [ComfyUI-0/STDERR] File "H:\SD-Zluda\SwarmUI\src\BuiltinExtensions\ComfyUIBackend\ExtraNodes\SwarmComfyCommon\SwarmSaveImageWS.py", line 48, in save_images 22:38:00.930 [Warning] [ComfyUI-0/STDERR] i = 255.0 * image.cpu().numpy() 22:38:00.931 [Warning] [ComfyUI-0/STDERR] RuntimeError: Numpy is not available 22:38:00.931 [Warning] [ComfyUI-0/STDERR]

[ornate elk]

Have you successfully generated an image with comfyui First?

[last tide]

No

Ill try that rn

That gives me the same error

[ornate elk]

Yea try to get comfyui work before using it in searm

[ornate elk]

Did you used the comfyui-start.bat ?

[last tide]

Yes

Got it working

[ornate elk]

nice

[last tide]

Installed comfyUI manager, went to ComfyUI\custom_nodes\ComfyUI-Manager\config.ini and wrote security_level = normal-, restarted server, went to "manager", "install pip packages" and wrote numpy==1.26.4 even tho it was already installed

Now trying to get it working on SwarmUI

🥳

Thanks for the help and support CS1o! Been amazing

[ornate elk]

perfect! no problem 🙂

[hasty orbit]

Suppose I had an instance that generates images all day, are there any automation tools out there that would allow me to swap out the prompt from a text file every hour?

[chrome tundra]

I got juggernautXL working with webui stable diffusion, but I saw this error right after loading the model for the first time. Should I be concerned? It has some issue with the dimensions of the NN which seems important!

While copying the parameter named "model.diffusion_model.middle_block.1.transformer_blocks.1.norm2.weight", whose dimensions in the model are torch.Size([1280]) and whose dimensions in the checkpoint are torch.Size([1280]), an exception occurred : ('Cannot copy out of meta tensor; no data!',).

[ruby yoke]

for which interface?

[hasty orbit]

I guess the default stable-diffusion-webui

[ruby yoke]

that would require something that would run inside webui that could edit the prompt - i'm not aware of anything that can do that

you'd probably need to run directly at the command line

[hasty orbit]

My temporary solution right now is using the "prompt from file or textbox" option and using a firefox plugin Automa to hit the "generate" button every hour

So every hour it will spit out 5 different images from 5 different prompts

[orchid vale]

hello, trying to figure out where i put in command lines. (--xformers --listen --api --no-half-vae --medvram --opt-split-attention --always-batch-cond-uncond). do i put them in webui user.bat or something else? thanks for the help

[oak jay]

Is there a way to use existing models folders with Forge?

[hasty orbit]

Windows or Linux?

If Windows it's in webui-users.bat, and you add the line:
set COMMANDLINE_ARGS=--xformers --listen --api --no-half-vae --medvram --opt-split-attention --always-batch-cond-uncond

If Linux, it's webui-user.sh, and you add the line like so:

[hushed prairie]

Hey guys, does anyone know what the difference between sd_xl_base_1.0.safetensors and sd_xl_base_1.0_0.9vae.safetensors is? They didnt mention it in the description

[ornate elk]

Only use xformers OR opt-sdp-attention, not both at the same time. And dont use --always-batch-cond-uncond

[karmic crown]

it would be best to use an internal SSD because it would work the fastest. you can move your models to an external driver later if you do not have enough space, but it is recommended that an external drive is also an SSD. I have M3 Pro with 1TB SSD + 1TB SDXC, and everything is on my SSD, since it is much faster than SDXC

[calm cairn]

Does anyone know whether SDXL has a minimum resolution required for inpainting and img2img?

[ornate elk]

Best would be 1024x1024 but 768x768 works too

[half island]

How do i properly install these "missing nodes"? As they are not ones i get from comfyui-manager

[last tide]

Anyone knows why my flux1-dev is taking an entire hour to generate a single 1024x1024 image?
Sdxl and dreamshaper8 are max 30 seconds.
(SwarmUI with comfyUI)

Also it says "1 current generation, 1 running (est 1m 3s)..."

Output also looks incredibly weird and blurred/pixellated

[ornate elk]

Which flux model do you use ?

Make sure you use this one:
https://huggingface.co/Comfy-Org/flux1-dev/blob/main/flux1-dev-fp8.safetensors
Its the fp8 flux dev but as 16gb variant

[last tide]

Im using the 22gb version, ill try the fp8 16gb one

[ornate elk]

Yea dont use the 22gb

[last tide]

Seems like it tries to allocate more cuda cores than i have, got 16gb vram, and its maxed out in task manager (windows)
15:37:40.598 [Warning] [ComfyUI-0/STDERR] File "H:\SD-Zluda\SwarmUI\dlbackend\comfy\ComfyUI\comfy\ldm\modules\attention.py", line 407, in attention_pytorch
15:37:40.598 [Warning] [ComfyUI-0/STDERR] out = torch.nn.functional.scaled_dot_product_attention(q, k, v, attn_mask=mask, dropout_p=0.0, is_causal=False)
15:37:40.599 [Warning] [ComfyUI-0/STDERR] torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 1.69 GiB. GPU
15:37:40.599 [Warning] [ComfyUI-0/STDERR]

"ComfyUI execution error: CUDA out of memory. Tried to allocate 1.69 GiB. GPU"

Must be something wrong with this model

Dreamshaper 8 works just fine

and now sd_xl looks weird aswell?

Looks like dreamshaper does work with regular swamui but not with comfyui? Weird

[zealous pagoda]

hi guys
I saw an interesting video from AI and wanted to start creating videos too, but after installing SD-CN-Animation nothing changed, help me solve this issue

[ornate elk]

I can send you a workflow for flux later

To test

[last tide]

Sure, would love to try

[ornate elk]

Here

[last tide]

Can someone ban this bot already?

[last tide]

Testing it out rn

Seems to be working perfect

[ornate elk]

Done

[tulip relic]

Hey guys new user here, what should i do?

[fiery bough]

I'm having an issue installing ComyUi manager. My Comfyui was coming up as a white screen after I clicked the update files in file manager so I uninstalled everything that had to do with comfyui and reinstalled it again. There may have been some files left behind because when I opened Comfy for the 1st time after reinstalling it, it still had my last prompt.

Now when I try to install the Ui Manager I get the error in the screen shot. Anyone know how to fix this? One thing that I notice is my terminal says _NVIDIA after windows protable, which in any other video I seen on installing manger, it does not say NVIDIA. I'm trying to run FLUX and I need to update my ComfyUi. Any help would be greatly appreciated Thank you!!

[ruby yoke]

it can't find your git install.

[last tide]

Its so slow compared to SwarmUI "generate", the workflow takes about 5-10min for a simple prompt. Yet "generate" does it in 2-3 min

Any reason for this?

[ornate elk]

same model?

[fiery bough]

how do i make it so it's able to find it? I thought I was just supposed to right click the link in the terminal. Did I miss a step?

[ruby yoke]

first question - did you install git?

[last tide]

Flux1-dev 6gb

[ornate elk]

6gb?

[last tide]

16gb, my bad

[ornate elk]

ah okay

and where does it take longer?

in swarm or comfyui

[last tide]

In the comfyUI workflow

[drifting pendant]

Hello there, @ornate elk Can you link me your tutorial for the command line args please ? I've lost my saved links 🥲

[ornate elk]

hey, here are all the Guides:
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides

[drifting pendant]

Thanks a lot !

[ornate elk]

np 🙂

[boreal vector]

the upscaler in extra tab doesnt work

it just gives a jpg of higher size

am i doing anything wrong?

[ornate elk]

you can try add --reserve-vram 1 to the comfyui-start.bat after the --auto-launch

[boreal vector]

https://huggingface.co/Kim2091/UltraSharp/tree/main

i decided to use a newer version of the upscaler but now it doesnt even show up

[ornate elk]

can you show an example?

[boreal vector]

that was with an older model

[boreal vector]

its not completely sfw but there is no nudity either

left is original right is upscaled

[ornate elk]

yep it didnt changed anything

[boreal vector]

yeah and now the new model doesnt even show up

i also restarted webui

[ornate elk]

what upscaler visiblity did you used?

[boreal vector]

upscaler 1 doesnt have any visibility option

how can i make Ultrasharp appear here

[ornate elk]

always use the .pth versions of upscaler

and the best upscale result is using highres fix or img2img sd upscale script

[boreal vector]

im doing it with highresfix

like this

[ornate elk]

your webui seems a bit outdated

and always set hires steps to 10

[boreal vector]

thats not my screenshot

i took it from a reddit comment about workflows

[ornate elk]

Ah okay

Yea you enable hires fix, set denois to 0.5 and hires steps to 10

[boreal vector]

my version is this

[boreal vector]

ig its doing something now

[boreal vector]

nope its the same

[ornate elk]

what does the cmd shows?

[boreal vector]

it says tiled upscaled 100%

there is no error in the process just the output

[ornate elk]

Then something isnt working

Can you share the whole cmd log after the upscaling

[fiery bough]

No, but now I do. I got it to work now. Thank you!! Now I just need to figure out to get flux on there lol

[ruby yoke]

you want to download a version of flux that'll fit in the Vram you have. how much vram do you have, and what GPU do you have? and are you running intel or AMD for the cpu?

[fiery bough]

I have a 7900x3d, a 4080, and 32gb of ram. I got flux working now and it seems to going fine. I got the DEV one. Takes about 40+ seconds to do the image.

[ruby yoke]

yup. dev will work fine in that 🙂

[wraith mirage]

do we have any tool to convert fp16 FLUX checkpoints into fp8 or GGUF-Q8?

[cloud flame]

What is the difference b/w Sd1.5 sdxl and flux? Rn I have automatif1111 is that just sd1.5?

Also can I use a pony Lora on a realistic checkpoint?

Does it generate smth meaning full?

Or do I have to find realistic loras for realistic checkpoints?

I'm currently aiming for that Korean idol look

With different poses, dresses etc

And I have found many of those loras but in pony

Wait can I use pony to generate semi realistic then?

Sry for so many questions

Similar to this one

[azure magnet]

If i import an image made with automatic1111 in comfyui should i have the same output image?

[fickle sorrel]

Can someone please explain to me what I'm doing wrong?

[fair oxide]

What does it mean by "negative values"

[ruby yoke]

you could just download the models that are already quantized for that

[ruby yoke]

when you desaturate, i.e. lower the number, your colors get less bright until finally all you have are shades of grey

[fair oxide]

The lowest it goes is 0

[ruby yoke]

i know. what is it you're trying to acomplish?

[fair oxide]

Some images look like color vomit and I want to tone it down

Do I get set it a little above 0?

[ruby yoke]

okay so i'd start by making sure cfg isn't too high, and adjusting your other values. what are you generating with?

[fair oxide]

Tponynai3

Cfg is at 7

[ruby yoke]

drop it to 4.

[fair oxide]

Ok will try

[lament fog]

I'm not an expert in pony and in fact have never used it. But I've seen many people generating realistic images with it, so I think the answer is yes. sd1.5 and sdxl are different models released by stability ai, in that order. The former was released about 2 years ago, and the XL release around july of '23. Flux just came out only weeks ago, and was released by an alternate company known as black forest labs. each new generation of model has shown pretty significant improvement in prompt adherence and possibly even other things such as better handling of rendering objects or people in the background (these have to do with the native resolutions mainly), but there are factors that keep the community attached to some of the older models. factors such as a wealth of community contributions (finetunes, loras), compatibility with extensions like controlnet, ipadapter, video creation extensions, etc

[fair oxide]

It looks much better

[ruby yoke]

excellent.

[fair oxide]

Still looks a little weird so I'm gunna try 3

[ruby yoke]

you might also play around with the samples and schedulers

[fair oxide]

3 fixed it

What even is cfg scale bruh

[ruby yoke]

very good. yeah, i think your screen shot is from an AI that thinks you're using photoshop, really.

[fair oxide]

Weird

[ruby yoke]

classifier free guidance scale. decides just how hard the sampler should hit.

[fair oxide]

So could the problem also potentially be fixed by using a different sampler?

It would just look different

[ruby yoke]

your samplers and schedulers are going to have a massive effect on the outcome, yes - and some don't work together well. and some don't work well with certain models but work well with others

they determine the math used

[boreal vector]

i dont see the error

the 8% i interrupted

[knotty turret]

should i use the command --medvram on webui forge i have 6gb vram

[jagged rampart]

What can you tell me about “search-and-recolor” in the “stability ai api”! It can be seen that the colors specified in the prompt are not uniform in the output for each reference image. Based on the above, is it possible that the color of the original image is affecting the output result, is this correct in your perception? In the future, we would like to make detailed color settings using color codes, but is there a setting that allows us to output the specified colors as they are?

[ruby yoke]

that's not how the AI works - you can prompt for a color, but it's not likely to give you the exact shade you're looking for - it's not photoshop, DaVinci Resolve, or Adobe Preimer

[hasty orbit]

Running into this issue where I'm generating images and then towards the last image out of nowhere a ton of swap memory gets used up and my webui instance just crashes

[ruby yoke]

maybe increase your swap space then?

(and make sure you're not running stuff like discord at the same time)

[hasty orbit]

Using 8gb vram, with --xformers --api --no-half-vae --medvram --opt-split-attention --always-batch-cond-uncond

[hasty orbit]

It's in a virtual machine on a hypervisor

[ruby yoke]

it still has swap space

[hasty orbit]

Completely seperate hardware and system

[ruby yoke]

what os?

[hasty orbit]

So the swap space is dependent on the physical RAM hardware I believe

Ubuntu

[ruby yoke]

https://help.ubuntu.com/community/SwapFaq this might be useful

[grim ermine]

im having some problems when i try to merge checkpoints and keep getting an error message

[hasty orbit]

Increased the swap memory to 8GB so fingers crossed

[ruby yoke]

can you post the error log here

[grim ermine]

*** Error loading/saving model file
Traceback (most recent call last):
File "E:\AI\stable-diffusion-webui\modules\ui_checkpoint_merger.py", line 21, in modelmerger
results = extras.run_modelmerger(*args)
File "E:\AI\stable-diffusion-webui\modules\extras.py", line 213, in run_modelmerger
theta_0[key] = theta_func2(a, b, multiplier)
File "E:\AI\stable-diffusion-webui\modules\extras.py", line 97, in weighted_sum
return ((1 - alpha) * theta0) + (alpha * theta1)
RuntimeError: [enforce fail at alloc_cpu.cpp:80] data. DefaultCPUAllocator: not enough memory: you tried to allocate 335544320000 bytes.

[ruby yoke]

not enough memory: you tried to allocate 335544320000 bytes. <--- that's an interesting number. where'd you get the models you're merging?

[jagged rampart]

Okay... Are there any future plans to update the system with additional functions, etc.?

[ruby yoke]

you could train a lora on very specific colors with specific color labels - and then use it when you generate

[boreal vector]

Can I get some help with this?

[jagged rampart]

Thank you for your answer!
I am envisioning an API implementation of stability ai for a model that has been finetuned by LORA, is this possible? We are using only models that have already been trained, so we are not able to consider this in our configuration process.

[ruby yoke]

i'm not sure if SAI would do that or not. You'd need to get in touch with Stability via their customer service email and discuss something special i think.

[high smelt]

what are the diff for these 5?

[wooden crown]

can anyone help me get started creating txt to img with comfy UI? I'm trying to generate images like the examples on this https://civitai.com/models/639937/boreal-fd-boring-reality-flux-dev-lora

[vocal burrow]

different levels of ease of use, customizability, compatibility, speed

• Automatic1111 / A1111's stable-diffusion-webui : https://github.com/AUTOMATIC1111/stable-diffusion-webui/ the most popular solution out there, biggest community and support
• ComfyUI : https://github.com/comfyanonymous/ComfyUI : second most popular solution, more tweakable but more complex to use, usually the fastest option available by a slight margin when compared to a correctly configured A1111
• SwarmUI (previously StableSwarm) : https://github.com/Stability-AI/StableSwarmUI : a reskin of some sort of the two above but in one package
• Fooocus : https://github.com/lllyasviel/Fooocus : probably one of the easiest solution out there for people who just want a big button to press and generate "pretty" stuff, but also one of the less tweakable.
• Forge-Webui : https://github.com/lllyasviel/stable-diffusion-webui-forge a fork of A1111 aiming to make things faster but break a lot of A1111 extension compatibility and it's development has been paused/abandoned for a while and just recently started back for who knows how long

[high smelt]

i see, thanks

[cloud flame]

damn thanks a ton

[warped terrace]

hello! Where is the best place to ask a question about ComfyUI and IPAdapterApply?
https://discordapp.com/channels/1002292111942635562/1204675216773619752/1285503094410444843

or else... is there a better more up-to-date upscaler for Pony or ComfyUI?

[unborn bramble]

What is a good amount of Buzz I should use for a SD1.5 CivitAI bounty for a 49 character pack? I'll collect and supply the images

Oh oops wrong channel

[brazen chasm]

Hello, i'm new to stable diffusion and wanted to set it up in my laptop, but in the webui-user.bat file i am having this issue. Can someone please help me with it, and also i'm using the .safetensors verson of stable diffusion instead of the .ckpt because i only found the .safetensors version.

[boreal vector]

python version

get 3.10.6

[brazen chasm]

I'm on python version 3.12.6

Will it not work on that version?

[boreal vector]

no

[brazen chasm]

Why tho

Is there a reson to it?

[boreal vector]

idk I also accidentaly updated it once and it didnt work

then downgraded

[brazen chasm]

So do i have to do everything from the begining?

[boreal vector]

no just uninstall this

[boreal vector]

and install this

[brazen chasm]

You sure?

[boreal vector]

yeah i also did once

[brazen chasm]

Okay, i'll check it out.

[boreal vector]

Your Python should be installed here
C:\Users\your username\AppData\Local\Programs\Python

[fringe mirage]

anyone can help me with this?

[boreal vector]

I tried two upscalers but they are not working.

Workflow-

txt2img tab
base, 512x512 resolution, 30 steps, 7 cfg
higresfix, 2x resolution, 10 steps, 0.5 denoise
extras tab
upscale
2x resolution,
model - SwinIR/UltraSharp

[brazen chasm]

Ummmm.... Now i'm facing a new issue here. I installed python 3.10.6 and remove the old 3.12.6 version and when went back to webui-user.bat this showed up, help.

[boreal vector]

did u restart pc?

[hasty orbit]

Increased the swap size and still crashing

[brazen chasm]

No

Is it necessary?

[boreal vector]

yeah

[fringe mirage]

anyone knows about PermissionError: [Errno 13] Permission denied: ?

[steady thistle]

I tried using ComfyUI Manager to install https://github.com/WASasquatch/was-node-suite-comfyui . When I restarted, it uninstalled torch and then did the same crash I was having previously. Here's my .bat that I'm using.

I carefully read about this but there were no depreciations and I didn't think anything was too crazy about it.

And this was my next error before rebuilding venv: Requirement already satisfied: numpy in d:\ai\comfyui_windows_portable\comfyui\venv\lib\site-packages (from torchvision==0.18.0) (2.1.1)
INFO: pip is looking at multiple versions of torchvision to determine which version is compatible with other requirements. This could take a while.
ERROR: Cannot install torch==2.4.0 and torchvision==0.18.0+cu121 because these package versions have conflicting dependencies.

The conflict is caused by:
The user requested torch==2.4.0
torchvision 0.18.0+cu121 depends on torch==2.3.0+cu121

To fix this you could try to:

1. loosen the range of package versions you've specified
2. remove package versions to allow pip to attempt to solve the dependency conflict

ERROR: ResolutionImpossible: for help visit https://pip.pypa.io/en/latest/topics/dependency-resolution/#dealing-with-dependency-conflicts
Running main.py with --auto-launch argument...

And trying to reinstall, here's the error. Apparently I have torch on my c:\ now and that's the conflict.

[brazen chasm]

@boreal vector i restarted my laptop and it didn't work, i'm really frustrated right now. What am i doing wrong here.

[boreal vector]

check your python installation

[brazen chasm]

Then what?

[boreal vector]

is it there

[boreal vector]

should be here

[brazen chasm]

It is there

Python310

[boreal vector]

did you change your bat file after upgrading python

[brazen chasm]

No

[boreal vector]

go to your sd folder and rename venv

to something else like ven1

[brazen chasm]

The only thing i did was type git pull in the notebook of the bat file when i had python 3.12.6

[boreal vector]

now relaunch web ui

[boreal vector]

that doesnt affect python its just for updating webui

[boreal vector]

it should redownload the missing files and you should be good

https://new.reddit.com/r/StableDiffusion/comments/12gnwxz/accidentally_uninstall_python_for_stable_diffusion/?utm_source=share&utm_medium=web2x&context=3

if u need more help follow this

[brazen chasm]

Are you talking about sd3 file?

[last tide]

😳 😳

[ornate elk]

You have to delete the venv folder, then relaunch the webui-user.bat

[steady thistle]

Sounds like this may be what I need too. Not sure tho.

[steady thistle]

Mine sounds like a very similar issue, but apparently I installed torch to the PATH install of Python rather than into the venv.

[ornate elk]

You installed kohya inside a OneDrive synced folder. Thats causing the permission issue

[ornate elk]

Comfyui does that

[fringe mirage]

C:\Stablediffusion\KohyaTraining\kohya_ss its here

[steady thistle]

Hmm, so now there's a conflict.

[ornate elk]

Look at your screenshot again

[fringe mirage]

yes?

[ornate elk]

There you see the path

Its inside OneDrive sync

So either disable OneDrive or move kohya to an other place

[fringe mirage]

A yea i moved from desktop to C

maybe thats the issue?

Do i need to install from scracth?

[ornate elk]

Nope both wouldn't work as OneDrive syncs the Desktop too

Make a new folder on C: and then install Kohya in it

Dont move it

Then it should work

[fringe mirage]

Tyvm

[steady thistle]

Did you see my issue above from a few minutes ago, trying to install WAS node suite?

[ornate elk]

Nope

[steady thistle]

Do you have thoughts on it? Looks like there's a conflict with installing a different torch version, but I also have no idea why it decided to uninstall torch with Manager restarting comfy

[ornate elk]

Open up a CMD and type
Pip cache purge
Then delete the venv folder and relaunch the webui-user.bat

[steady thistle]

Was this for me? Because it seems like a good idea.

[ornate elk]

Nope

But you can try

Pip cache purge doesn't hurt

[steady thistle]

Okay

[brazen chasm]

The venv file is in the trash can, should i completely remove it for it to work?

[ornate elk]

Yep

You don't need it

It will get rebuild after launching the webui-user.bat

But now with the correct python version

[steady thistle]

I tried this, then I did venv\Scripts\activate but system cannot find path specified.

python got screwed up?

[ornate elk]

You don't need to run it inside the venv

[steady thistle]

I did the purge, but then went to try and rebuild venv but I couldn't get into the environment

then I remembered to do python -m venv venv sorry, painfully bad at this

I then did pip install -r requirements.txt and it's downloading torch now, but it didn't go to the right version I'm afraid. Gonna have a conflict.

unless.... I wonder if my issue is because requirements.txt is out of date..?

Hmm, yeah, I could put versions in the requirements.txt

[brazen chasm]

Brothers, i'm still having the same problem and now i'm thinking about killing myself, there has to be a way to solve this issue.

[steady thistle]

AssertionError: Torch not compiled with CUDA enabled

[ornate elk]

Install python 3.10.11 64bit

Then relaunch the webui-user.bat

Currently it doesn't find your python so installing the latest compatible one should fix it.
Then if you have a venv folder again, delete it before relaunching

[steady thistle]

I'm trying pip install torch==2.4.0 torchvision==0.18.0 torchaudio==2.4.0 --index-url https://download.pytorch.org/whl/cu121 again. I have no idea why it didn't work before but hopefully the pip purge helped.

I made sure requirements.txt matches the version in the .bat so that if it runs through the requirements again it won't start uninstalling all my work

[boreal vector]

no luck

<- base upscale ->

[steady thistle]

@ornate elk I don't know what to do at all, I haven't faced this error before today. I'm stumped.

I don't know how to safely remove package versions from PATH

[fickle sorrel]

Can someone please help me get this to work?

[fringe mirage]

still the same error

[spring spear]

does stability still have the colab notebooks?

[upper moth]

I'm used to getting image info pretty easily on Automatic1111, but I have recently moved to ComfyUI, is there a way to get PNG info there too?
Or with some outside tool that woudnt require me to launch A1111 (cant get it to work on my PC)

[solemn moon]

all the info is stored in every output, drop the image back into comfy and it will give you the setup complete

[ornate elk]

Okay then edit the folders permission and give yourself every permission

[ornate elk]

Can you try an other upscaler? Like latent bicubic ?

[harsh timber]

help plz

[ornate elk]

What's your GPU?

[harsh timber]

amd

7900 xt

i know have this line, but this uses my cpu instead of gpu if im right

[ornate elk]

Yea because you used the nvidia webui version

[harsh timber]

is there a amd version?

i just followed a very simple guide on yt

[ornate elk]

You need to use the AMD Webui with Zluda Guide from the pinned messages

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

[harsh timber]

Do i need to delete my stable diffusion map?

and restart all over again?

[ornate elk]

Yes

[harsh timber]

okay

i will try

[ornate elk]

For any questions feel free to ask me

I'm using a 7900xtx

[harsh timber]

okay thanks

do i still need to use sd 1.4 script, gfpgan and git?

[ornate elk]

Only python 3.10.11 64bit and Git

Is needed to start the guide

[harsh timber]

i already have those installed

this one right?

[magic lintel]

Does anyone know if 2 minutes to generate an image is normal on an rtx 3070?

[harsh timber]

@ornate elk i am stuck at this part

[ornate elk]

With which settings

[ornate elk]

You can ignore step 3

[harsh timber]

im at step 4 now

i don't understand\

[magic lintel]

https://i.gyazo.com/c4081bbd67eca3c3e954601a15d702bf.png

also its making hideous monstocities

like whats in the picture

[harsh timber]

yeah im right there yes

[ornate elk]

Its and sdxl model and pony too. And your generating at a very large resolution.
But first show what's inside your webui-user.bat?

[harsh timber]

but i don't understand what they mean with this

[magic lintel]

i'm so sorry. How do i get inside to look at that?

[harsh timber]

do i have to copy word C:ZLUDA and %HIP_PATH%bin

[ornate elk]

You need to create these as 2 new entrys

[ornate elk]

Right click and edit it

[harsh timber]

i can't edit or do anything at system variables

[magic lintel]

https://i.gyazo.com/df51f1ccd5797af4c3517557ed755bf5.png

https://i.gyazo.com/53fd2dc7a6ea093d541c5605af4fae8b.png

[magic lintel]

also what does it mean if images come out like this?

[ornate elk]

Thats the webui.bat not the webui-user.bat

[ornate elk]

The VAE is missing

[ornate elk]

Can you screenshot it?

[magic lintel]

sorry it says this inside

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=
git pull
call webui.bat

[ornate elk]

Okay your missing the performance args.
At the line Commandline_args=
You add: --xformers --medvram-sdxl --no-half-vae
Then save and relaunch the webui-user.bat

[magic lintel]

like so?

https://i.gyazo.com/d79f75b04cbd56eb6b33c193908f31c5.png

[ornate elk]

Yep

[magic lintel]

this will improve generation time?

okay looks like the xformers are installing

[ornate elk]

Yep should be really fast now

[magic lintel]

hmmm its still not

says 30 minutes even now

[ornate elk]

Show your settings again?

And dont generate at 2k

[magic lintel]

https://i.gyazo.com/f3a9b04e5a8354a264ecc0e76af7676b.png

[ornate elk]

That makes no sense. Do you run any game or Wallpaper engine in the background?

Also use sampler Euler a or DPM++2m

[magic lintel]

no just chrome

[ornate elk]

And 30 steps

Dont use restart sampler

Thats slow

[magic lintel]

got this too when i tried to apply the VAE

https://i.gyazo.com/4dd20d2426ba90da50b2aa39fe186d17.png

[ornate elk]

Which setting did you changed ?

Only the vae ?

Then its okay

[magic lintel]

yeah tried to apply it

and this happened

woah thats fine?

[ornate elk]

Yea thats normal

xD

[magic lintel]

lol

https://gyazo.com/ddd2e902761f19a323d0b15a1e39478f

this eta s normal?

says 5 or 6 minutes

[ornate elk]

Can you try a 1024x1024

Better to compare for the speed

I'm AFK rq

[magic lintel]

https://i.gyazo.com/f76f740f55f1e9d22b8bfd3c1645260e.png

still says 6 minutes

so crazy it was working totally fine yesterdauy

a friend said to clean install after i was having issues with the VAE

since reinstall its been so slow

okay the VAE worked

now i just need my regular speed back

im cool with 1-2 minutes for a high quality image, but 5-6 feels kind of weird

and i just put in a new card

[ornate elk]

Yea 6mins isnt right

Make sure you always launch the webui-user.bat and not the webui.bat

[magic lintel]

okay so i got it back to being fast but now faces are fucked up

[magic lintel]

i just installed adetailer too

im sorry about all these questions

for every fix there seems to be a new problem

does this face look fine considering i installed adetailer?

this ones better cause i upped resolution i think

sorry if its rude to post these dunno if this is the right chat

will delete if its cluttering

[ornate elk]

Its okay to post them if its not nsfw

[ornate elk]

And how?

[magic lintel]

lowered resolution like you suggested

[magic lintel]

also restarted and made sure to use webui-user

[magic lintel]

do these look good to you?

just made this

took about a 2 minutes to make this

thats prolly normal right?

2 minutes with adetailer

https://i.gyazo.com/43b1c6c1e632e2b401e109345e29b9a2.png

with these settings

[ornate elk]

looks good

@harsh timber you still need help?

[harsh timber]

oops sorry, i didn't see your previous message, no im gonna continue another time, but thanks anyway

[magic lintel]

whats it mean if i get shit like this?

tried to make this

https://civitai.com/images/25626865

it says its for flux right?

[loud pond]

@ornate elk have you got a tutorial on how to download flux?

[ornate elk]

alright, let me know then ^^

[ornate elk]

whats your gpu and do you have a webui already installed like comfyui or forge =

[ornate elk]

flux doesnt work in auto1111 currently

[loud pond]

amd radeon 480, id like comfyui but i havent got it installed yet

[ornate elk]

with 4gb ?

vram

[loud pond]

no 8gb

[ornate elk]

hmmm

i dont think it can run flux :/

[loud pond]

ffs

thanks tho

[ornate elk]

but your gpu is supported by ZLUDA it seams which means you could run sdxl models

[loud pond]

yeah i have SD already

but apparently flux is more relistic

realistic

[ornate elk]

yea but we have to weight for AMD to support bitsandbytes on windows. so we can use the smaller nf4 or gguf models

[magic lintel]

ahhh darn

you use auto1111

[magic lintel]

also thank you so much for all your help again dunno if i said that

[ornate elk]

no problem 🙂

[boreal vector]

they not here

[ornate elk]

then try Nearest

[boreal vector]

it didnt even show up on cmd

[magic lintel]

can i have comfyui and a1 on the same comp? I know that sounds dumb but like they don't conflict or anything right?

[ornate elk]

try updating your extensions and then relaunch, make sure the webui is whitelisted in any browser adblocker.
And if its still not working try deleting the config.json and the ui-config.json and then relaunch

[ornate elk]

yes you can have them both without problems

i have all webuis in one Folder

[magic lintel]

you think i can make decent stuff on FLUX with 8gb of vram?

[ornate elk]

yes but for that you need comfyui or Forge webui. and the nf4 v2 flux model

[boreal vector]

doesnt work

[cloud monolith]

i was using it jsut fine, was making charcter art. btu then something happened an dnow it just makes blobs of color

how do i fix it. as now no matter what i do its doing this

[ornate elk]

can you show an example with the settings you used?

[ornate elk]

then you could try a clean install with my guide and test again. and dont install any extensions at first

[cloud monolith]

the lora for the chaaracter i want was workign perfectly fine last night witht he exact same promt

[ornate elk]

wrong resolution for a sdxl/pony model

these need a highre base resoltuion

[cloud monolith]

then why is it still doing it after making the res bigger

and why did ti work perfectly fine last night

[ornate elk]

looking at the screenshot again it seems that your negative prompt is not recognized

make sure the webui is whitelisted in any browser adblocker

[cloud monolith]

ad blocker hasnt chnaged since it woreked

[ornate elk]

your sure? they update their blocklists in the background

[cloud monolith]

yup.

[ornate elk]

the 0/0 in your negatives is caused by something blocking it or not updating/sending it to the cmd

[cloud monolith]

so how do i fix it?

[ornate elk]

first thing i would try is to relaunch, or to whitelist the webui in any browser adblocker xD

[cloud monolith]

already did

maultiple times

[ornate elk]

can you show the cmd log, maybe we can see an issue there

[ornate elk]

i mean the full cmd log

also why do you use 150 steps with an sdxl/pony model? ^^ 30 would be enough

[cloud monolith]

is that not a full llog?

also 30 looks worse

[ornate elk]

Nope

Thats just the end of the log

You can relaunch and then generate one image and then copy the whole cmd log

[ornate elk]

Okay 5 things I can see:

• Your webui is not updated since Dec 23 (9 months outdated).
  And since it was not installed the right way it can't be updated easily.
  Your on 1.7.0 while latest is 1.10.1
• Your not using the recommended performance Arguments for your gpu, meaning you could generate much faster.
• You have Roop Extension installed which is deprecated and causes errors in your log.
• Your using Lycoris extension, but its not needed anymore, even on 1.7.0.
• you have other extensions that cause a flood of requirement checks at the start.

What I would recommend is a clean reinstall of the webui.
Its hard to fix outdated stuff.
But if you want to keep using 1.7.0 then you should remove Lycoris and roop and any other extensions you don't need and after that delete the venv folder and relaunch.
For a better performance you should checkout the install guides in the pinned messages. They show the right args for your GPU vram.

[cloud monolith]

how do i find those?

[ornate elk]

The extensions or guides ?

[cloud monolith]

both

but for starters the guide to do it properly

[ornate elk]

In the pinned messages of this channel or directly here:
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides

[cloud monolith]

but.. i followed that guide when i did it tho o.O

[ornate elk]

But the console says commit hash none. Means it was not installed via Git 😮

Or maybe it was moved, that breaks it too

Also the performance args are not in your webui-user.bat

Or dont get recognised

[hasty orbit]

I'm still running into the issue where even after increasing swap memory, the application is still crashing

Should I just keep increasing memory (not vram)? The recommended is 32GB but I can increase it further

I'm starting to think there's something wrong with Stable Diffusion not releasing memory

[ruby yoke]

it wouldn't be stable not releasing, it would be the interface you're running stable in. and that could be the case.

[ornate elk]

Do you try sdxl or pony models?

Then you need a swap of at least 32gb

[cloud monolith]

uhhh it says for nvidia thers no gpu

lioek when i cloick thte option for that int he guides it dosnt do anything

[hasty orbit]

I use both

[hasty orbit]

I'll bump the swap to 32gb then

What about regular memory? Leave that at 32?

[ornate elk]

Your correct, I broke my link it seems. Let me link you directly to it.
https://github.com/CS1o/Stable-Diffusion-Info/wiki/Webui-Installation-Guides#nvidia-automatic1111-webui-stable-diffusion-webui

[cloud monolith]

ty

[ornate elk]

Np, will fix the link tomorrow

[ornate elk]

Do you use nvidia or amd ?

[hasty orbit]

Nvidia

[ornate elk]

A friend of mine used a 6800xt on Linux and had massive RAM issues.
Because on Linux the webui seems to use or need more RAM.
He increased the swap to 32gb, that helped but in the end he bought 32gb additional RAM

[hasty orbit]

I'm also referring to non-vram

[ornate elk]

Yep ik

[hasty orbit]

I'll keep you posted, and thank you for the assistance so far!

[ornate elk]

Np, also when does it crash?

At upscaling ? Or launching ?

Make sure you also use --xformers

[hasty orbit]

Yup using --xformers, so it crashes towards the middle or end of generating a small batch of 3 images.

Most likely during upscaling

I'll click generate and it will generate the 3 images no problem, But there are times by chance once it nears the end of the 2nd image or end of the 3rd image it just crashes.

[thin cave]

I don't know what i'm doing wrong https://i.imgur.com/BRDv1N9.png

[ornate elk]

Make sure to always set the hires steps to 10

If its on 0 it will most likely crash

And then there is the tiled VAE extension. Which can prevent crashing when upscaling too

[hasty orbit]

Should I try this first before changing the swap to 32gb?

[ornate elk]

You did everything correctly. But zluda found your CPUs integrated GPU before your normal GPU xD

So you need to disabled Radeon TM Graphics in the device manager

Under display adapters

[ornate elk]

Sure, maybe it will fix it

[brazen chasm]

Help with this issue please in webui-user.bat

Like i've done everything that's required me to do but there's always seem to be a problem

[ornate elk]

Make sure your nvidia GPU driver is updated

[brazen chasm]

And?....

[ornate elk]

And then relaunch.
And maybe restart the PC too.

[brazen chasm]

The webui-user.bat file?

[ornate elk]

Everything else looks correct in the screenshot

[thin cave]

sucess thanks my man

[ornate elk]

Perfect np

[steady thistle]

@ornate elk I'm trying some stuff to get my ComfyUI running again, but it's very challenging, and I don't understand why it keeps breaking. Is there something in my system that's going wrong, do you think?

I was careful to research https://github.com/WASasquatch/was-node-suite-comfyui and then restarting through Manager it uninstalled torch and started reinstalling it. I have no idea why it did that.

[thin cave]

I believe i've hit another wall https://i.imgur.com/B2IhPcA.png

[fickle sorrel]

Can someone help me please? Why am I just getting noise?

[ornate elk]

What's your GPU ?

[ornate elk]

Havr you tried an other sampler ?

[thin cave]

6900xt

[ornate elk]

Okay so you have hip SDK 6.1 installed ?
Restarted the PC too?
And how do your environment path settings look?

[thin cave]

yes yes and yes

[ornate elk]

The last one is not a yes or no question xD

[thin cave]

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

took me a sec

[ornate elk]

Thats wrong

Add them here, make 2 new entrys

And remove the wrong ones you added

[thin cave]

does this look correct? https://i.imgur.com/QGk9Cl7.png

[ornate elk]

No

Look at my screenshot

Find the left Path entry in the red square. Then click edit

[thin cave]

https://i.imgur.com/iFk0lhu.png

there are none

[ornate elk]

You removed it lol

In your other screenshot it was there

[thin cave]

shit

[ornate elk]

Then you need to recreate it

[thin cave]

https://i.imgur.com/Mwm0cND.png

[ornate elk]

Can you go into edit mode again?

And screenshot it

[thin cave]

https://i.imgur.com/F8dclqV.png

[ornate elk]

Why does it look so strange

Normaly you should get a list view

You also need the first 4 entrys here in Path, to not get windows problems

But you need to go into the right edit mode

[magic lintel]

@ornate elk Sorry but i have a very dumb question. I just downloaded comfyui and followed all the instructions, so its working but it only opens in its own program what do i press to open it up in browser like stable diffusion?

https://i.gyazo.com/d94bf7a6daed3ea65faa6e4e3267c99c.png

[ornate elk]

Seems like you installed it via stability matrix or pinokio.
You can just copy the URL you see at the top right into the browser

[magic lintel]

yes the tutorial advised pinokio for easiest install

thats cool you'd say?

or is there a better way?

[thin cave]

this is the edit mode that I have https://i.imgur.com/aq9s3xo.png

[cloud monolith]

i just reinstalled stable diffsuion. follwoed the steps in guide exactly. can anyon ehelp me confirm its all working right before i go further?

[ornate elk]

Sure

[cloud monolith]

just finished its download and i have the stable diffusion thing open in my brioweser

[ornate elk]

There is a // instead of a / at the zluda path

[cloud monolith]

does this look right?

[ornate elk]

Check the bottom of the webui

[cloud monolith]

for what?

[ornate elk]

There you see the version numbers

[cloud monolith]

version: v1.10.1  •  python: 3.10.11  •  torch: 2.1.2+cu121  •  xformers: 0.0.23.post1  •  gradio: 3.41.2  •  checkpoint:

[ornate elk]

Yep looks good

[thin cave]

does this look right? https://i.imgur.com/CQo0rQD.png

[cloud monolith]

it using my Nvidea gfx right? and not my laptops intigradted gpu?

[ornate elk]

Now your missing the : after the C

[ornate elk]

Yes

[thin cave]

this looks correct then? https://i.imgur.com/Yuk5nmD.png

[cloud monolith]

ty for all the help

https://tenor.com/view/hug-warm-hug-depressed-hug-gif-4585064738068342394

[ornate elk]

No problem 🙂

[ornate elk]

Btw what's your GPU ?

[cloud monolith]

nothing super fancy. just an nvidia geforce rtx 4080 laptop gpu

[ornate elk]

Ah okay, but thats a nice GPU for stable diffusion

Only a bit sad that the laptop version has 12 instead of 16gb vram

[cloud monolith]

yeahh

oh.. jutst relized. I have to redownload all my stuff again >.<

43 models and like 320 loras

😭

[ornate elk]

You can move the models over

Or did you deleted them already ?

[cloud monolith]

i deleetd and wiped it all before reinstal. did the same gor graphcisdrivers and evrything jsut to be 100% sure

[ornate elk]

Oh okay

[cloud monolith]

i wanted to be completly sure, it wasnt something else thart was the problem

[ornate elk]

Yea thats good now it should be as fast and clean as possible

Lycoris can be dropped into the lora folder and instead of roop you can use the Reactor extension

[cloud monolith]

dumb question. it is the wbui-user batch file i use to open table diffsuin every time right?

[ornate elk]

Yep and only the webui-user.bat

@thin cave
Let me know how it goes.
But I need to sleep now. Can help you tomorrow if its not working.

[thin cave]

seems to be stuck when attempting to generate https://i.imgur.com/gvZzD1I.png

[hasty orbit]

After changing high res steps to 10, it looks like it crashed again

Gonna change swap to 32gb

It almost looks like it isn't releasing memory from swap

[kindred pulsar]

It says that five tracks are pending, but it's three. The three have been pending for 2 hours.

[steady thistle]

@ornate elk I got it going, so far so good! Thank you again for your help.