#🤝｜tech-support

1 messages · Page 42 of 1

ornate elk
#

If you follow the guide you can still use the Directml webui + the zluda one separately

paper crag
#

I'll see how it goes.... sigh, i honestly don't know what happenend, just a week ago i was doing the very same generations, with 4 images at once and it was working fine

ornate elk
#

Zluda doesn't have these problems and the speed increase is enormous.
Also the vram management is much better

#

Didn't found a downside as since the 1.8.0 update even sdxl and IP-Adapter works

#

Speed increase is between x2 and x4

dense obsidian
#

I can vouch for that, and you use less board power = saves more money

umbral plover
#

I'm trying to use playground model in webui-forge but there seems to be some kinda of problem because i get only noised images, no errors in console

image.png
elder turtle
umbral plover
paper crag
#

question, for step 2, when i download the zluda zip file, do i need just 'ZLUDA-windows-amd64.zip' alone or all files ?

inland zephyr
#

It's probably better to use Hires fix to improve details than to upscale, right?

worn frigate
#

I think so.

#

Upscale will just upscale not fix it actually.

inland zephyr
#

Here comes my hero

ornate elk
ornate elk
inland zephyr
#

Wait hires is better than using the extras?

worn frigate
#

Maybe use adetailer or detail tweaker lora if want to add details like u mentioned.

inland zephyr
ornate elk
worn frigate
ornate elk
#

Hires fix in txt2img and SD upscale script in img2img are better

inland zephyr
#

Upscale script eh

#

I'll have a look at that

sterile garden
#

Unfortunately, it seems the extension didn't help me. Let me know if you have any other ideas. Thank you

00002-506080449.png
paper crag
ornate elk
paper crag
#

so i just press new and add that text?

ornate elk
patent carbon
#

Hi can someone help me, I installed python and stable diffusion but I get this error whenever I try to run the webui-user.bat

Screenshot_360.png
paper crag
ornate elk
#

It tried to use the Igpu of your Ryzen cpu

#

Instead of your gpu

ornate elk
#

And then relaunch

ornate elk
paper crag
ornate elk
#

But i would recommend to reinstall the Extensions again

#

And not move them over

paper crag
#

okay, the extenssion i'll do them again

ornate elk
#

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

#

In the taskmanager under performance tab, then on GPU

#

Ah okay

#

Then you only need to add
--xformers --no-half
To the line where --medvram is

#

Then the embeddings should work and it also should be faster

#

No problem 🙂

upbeat crane
#

have not had the greatest of luck so far pushing deforum through WebUI Forge. i was able to do a nice 20 second render on my first try using XL Canny and XL depth, zero issues, used sampler scheduling and four LORAS, not even a hint of an OOM. but for whatever reason now, i only get controlnet errors. made no changes to the system whatsoever, just started out of the blue. here is the entry:

#

*** Error running process_before_every_sampling: D:\AI\stable-diffusion-webui-forge\stable-diffusion-webui-forge\extensions-builtin\sd_forge_controlnet\scripts\controlnet.py
Traceback (most recent call last):
File "D:\AI\stable-diffusion-webui-forge\stable-diffusion-webui-forge\modules\scripts.py", line 835, in process_before_every_sampling
script.process_before_every_sampling(p, *script_args, **kwargs)

ornate elk
#

Nice no Problem 🙂

upbeat crane
#

think i'm going to do one last (and 4th) install of Forge, but go with the windows-installer package this time. prob wont do it for a while tho, so if anyone has any ideas that might help me would be much appreciated. cheers and happy blursday

vast crest
#

Hi, I have another problem, I saw that the stable diffusion version 18 is there. How do you think I can pass all the files I installed on the old version to this new one?

upbeat crane
#

i also tried moving my controlnets physically over to the forge installation and removed the --controlnet-dir out of webuser-ui just to test, it did not help, so i set it back like it was

#

just make a folder and toss em in, i say. controlnets, checkpoints, etc. i usually just screenshot my extensions and such then just go back and install those again after new version

#

or you could name your stable-diffusion-webui to stable-diffusion-webui_old, then just copy/paste the directories back over into your freshie

ornate elk
#

To update the webui you need to add git pull into an empty row in the webui-user.bat before above webui.bat

vast crest
#

this is what I have in my webui user: @echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --xformers --no-half-vae --autolaunch

call webui.bat

#

What should I do to get it updated without having to move files from one folder to another?

ornate elk
#

There you can add
git pull

#

It needs to be in its own row

vast crest
#

like this?:

git pull

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --xformers --no-half-vae --autolaunch

call webui.bat

#

and you should update to the new version, right?

ornate elk
#

If you launch it now it should update

vast crest
#

Does it update without me losing all the files I have?

upbeat crane
#

u wont lose your loras/checkpoints/models etc ur golden

#

cant wait for stable3, forge teased me so hard oof

vast crest
#

stable3?

upbeat crane
#

stable diffusion 3

vast crest
#

but we are at version 18. what is this stable3?

upbeat crane
#

ill keep using forge but i was just hoping it was gunna be my savior for pushing sdxl models through deforum

ornate elk
#

1.8.0 is the webui version

#

Not the model version

vast crest
#

ahhh,ok

ornate elk
#

SD 3 is a new model

vast crest
#

strange thing, it doesn't update anything at all XDDD

image.png
upbeat crane
#

i trust that a lot of these kinks will work out of webui forge as well, its really new

vast crest
#

today is not the day hahahaha

ornate elk
#

It will show why its not updating

vast crest
ornate elk
#

Ahh okay

#

You installed the webui the wrong way by downloading it as zip file

vast crest
#

ah, so how was I supposed to download it?

ornate elk
vast crest
#

so I create a new folder, use the CMD command, insert the link you put in the tutorial, follow everything, then I will have to manually insert all my files apart XDDDD

upbeat crane
#

you'll have to do a little maintenance now to get it installed correctly, which will save you headaches down the road

#

thats the beauty of installing via git clone, cuz later you can just git pull and it updates everything while leaving your downloaded content in place

#

also you will have more and more issues down the road if you dont do it

vast crest
#

I have a lot of work ahead of me, I have so many files to move. You would never guess how many GB I have to move XDDD

upbeat crane
#

it's not that bad honestly, takes me about 5 min

#

just a bunch of clickin and draggin

vast crest
#

I have 1TB of files to move 🙂

#

I think it will take me more than 5 minutes even if they are in the same SSD XD

upbeat crane
#

they dont weigh anything lol

vast crest
#

Positive thought XDDD

upbeat crane
#

seriously tho what all do you have to move, but these folders embeddings/controlnets/loras/checkpoints/VAE and any upscalers you may have in your RSGRAN and ESRGAN folders. just take a quick picture of your installed extensions, then you can go through super quick and reinstall those, bing bang BOOM

onyx solar
#

Moving files on the same drive shouldn't take any time at all...it doesn't read/write the files again, just changes their indexed location.

upbeat crane
#

ctrl-a copy ALL and drag into the new folder, do that like 6 times and your done

#

the files dont even copy they move, since they are on the same drive, but i digress

#

good luck, you'll feel better when your done and also knowing you have the program installed right using git from now on

#

make ur life easier moving forward, promise

#

also, this wont be the last time you are doing full reinstallations of stable diffusion

vast crest
#

done, everything perfect finally

upbeat crane
#

well done

vast crest
#

the next version, I want to try the git pull command and see if it updates me, so I'll know I installed correctly 🙂

ornate elk
#

You can also just link them

#

But if the new and the old webui are on the same drive you should move the models

#

That also shouldnt take long

vast crest
#

I can finally continue my creation of a female Vegeta 🙂

#

After creating Goku in the Female version 🙂

upbeat crane
#

that's the spirit

vast crest
#

Here's my female Vegeta 🙂

upbeat crane
#

i was expecting something completely different

vast crest
#

Not bad 🙂

woven canopy
#

How much vram is it normal for SD to use when you're not generating anything?

left iron
#

Why is it that on some UIs (UIUX, Forge), every so often the Generate button will just break and force me to refresh before I can generate anything

#

I don't see any debug output pointing to why, nor am I clear on what extension might be responsible

upbeat crane
#

prolly adetailer, that little bastard

left iron
#

oh man fr?
cause i tried fiddling with the keyboard shortcut to gen instead, gen forever, etc and nothing
let me try running for a while without adetailer

#

or at least without activating it

#

at least i figured out you can copy paste infotext

upbeat crane
#

adetrailer breaks deforum installs sometimes and causes all kinds of shit

#

been going through extensions too and troubleshooting

#

but my issues persist even with everything disabled

#

i think forge is just buggy rn, nothing but issues since the second i touched it

#

(and also fell in love with it at the same time)

#

forge is SOOOOO good for memory optimization and such

#

but i'm having a HELL of a time with it

vast crest
#

Anyway, guys, but is it normal for the image creation preview in: img2img to be so small? I can barely understand if it's creating a good image for me or not

upbeat crane
#

yeah mines been like that too lately, i just CTRL + or mouse wheel up

#

just one or two clicks will help

vast crest
#

I'm creating a sequence, which once finished, will look like an animation 🙂

ornate elk
ornate elk
vast crest
upbeat crane
#

you should try deforum

ornate elk
vast crest
#

However I always had the large preview without needing to zoom. Do I need to change some parameters in the settings?

vast crest
woven canopy
# ornate elk 2-3gb max

Ok, mine is using 8.5gb as soon as you boot up a1111 😵 I have xformers ruining, and I downloaded an extension with a "free GPU" button but it doesn't go below 8gb. What causes that kind of issue?

ornate elk
viral fox
#

hi All, Im brand new to this, but not tech. I have Stable diffusion SDXL installed and running on windows 11, with ComfyUI manager v2.9 (24-03-07). I am working through a YouTube video "L2: Cool Text 2 Image Trick in ComfyUI - Comfy Academy" and in it one of the nodes he adds is a ControlNet PreProcessor. Going through manager in ComfyUI the only preprocessor under the ControlNet menu is "AIO Aux PreProcessor" I have spent hours searching for this but cant find it anywhere. any ideas what I am doing wrong apart from being exceedingly thick in a subject I don't really know?

image.png
#

The other I cant find is

image.png
upbeat crane
#

you have the preprocessor for the controlnet Depth installed and working now you prob just need the actual controlnet itself

#

which is exactly the controlnet im troubleshooting right now

#

find out where your controlnets get installed in comfyui and look for it

#

this one, i think you might have to actually rename, because it wants to just call itself diffusion_pytorch_model.fp16.bin

#

so if u DONT have this CN downloaded and placed in your ComfyUI controlnet folder, grab this one that is 2.5gb, rename it and put it there

#

def no comfy user here, i gave it a go for a bit.....i probably need to start going back and learning it

#

thats weird that it shows as a preprocessor but you dont see it in the folder

#

ps: im only assuming thats the one you needed since that was the selected pre-processor, good luck

#

my sdxl depth isnt working in webui forge right now yeek

ornate elk
upbeat crane
#

well dude that might have just solved my issue @ornate elk lol. im using Forge, which was written by illyasviel, and im trying to use a Depth controlnet that he didnt actually create. sick

#

what a great lead, ty

woven canopy
ornate elk
woven canopy
#

Huh

upbeat crane
#

nope damn that is the exact same one i have "diffusers_xl_depth_full.safetensors" crap

ornate elk
#

from sai for exampel, or t2i

woven canopy
ornate elk
#

it breaks more than it helps

#

if you really have 2 gpus, it may catched the wrong one.

upbeat crane
#

digging through github and seeing tons of people left and right having controlnet issues with forge

#

fair enough, gotta move onto something else

#

back to learning comfy

ornate elk
upbeat crane
#

i have zero earthly idea after 3 installations why mine will NOT do the Forge integrated controlnets thing

#

folders are all empty, drop downs empty, git pull shows everything up to date, etc.,tried different folders, drives, etc.

#

tried under miniconda, without conda. good times 😉

#

finally with your help got my current CNs to start populating into Forge

#

and JUST like im seeing a ton of people saying

#

they did ONE successfully run using controlnets with this way, then after that first time

#

it breaks and nobody has a solution beyond that point, lol

#

i can only laugh tho, cuz i spent a ton of time prompting, ended up creating one of the best Deforum animations i have ever done, using Canny and Depth in Forge, using 4 different loras and even sampler scheduling. it pulled it off flawlessly and INSANELY FAST with no memory issues in sight. then after that render. goose egg

viral fox
upbeat crane
#

shit theres like 4-5 of each one but just built by different people

ornate elk
#

there are mutliple that do the same

upbeat crane
#

canny, depth, openpose, or dwpose, HED, maybe sketch

#

not much luck for me tho historically speaking with diffuser models, so i might take ole Sarge's out for a spin

viral fox
#

than , having downloaded I just copy inyto my controlnetUI/models/Controlnet folder?

#

im just getting into this. so its all new. im constantly searching for terms. cfg, seed, etc

upbeat crane
#

i just say "thingamabob" a lot

viral fox
#

whirrlygig

upbeat crane
#

doowhickey dealie-whopper

viral fox
#

ahh the old doowhicly isnt connected to the thingymabob so the whirlygig aint working

upbeat crane
#

ive decided to just download FOUR different depth controlnets, load them all into forge and test each one

#

like throwing spaghetti at the wall type situation

viral fox
#

hoping one is al dente

upbeat crane
#

exactly

viral fox
#

one i like, its a bit Mil, but im ex mil so. "Grenades are like quality street, make sure everyone gets a bir"

upbeat crane
#

niceee, army brat and ex navy here. cheers

viral fox
#

oxy moron an aemy matlow

upbeat crane
#

i would have joined the army but i didnt want to end up like my dad

viral fox
#

army

#

hahaha

upbeat crane
#

hehe

viral fox
#

so much to say, but well its a public forum

glad fiber
#

I am trying to train a lora, but i have noticed that all saved models have the same exact size and the generated image does not look like what i am trying to train

image.png 📎 train.json
#

how can i fix it?

#

and lloks like that if i increase the network rank dimension the model increases in size

viral fox
#

well that didnt work 😦 stoll no ControlNet PreProcessor

glad fiber
upbeat crane
#

yeah in comfy you have to actually find and download each CN pre-processor individually huh

#

i never got into comfy deep enough to do any CN testing

viral fox
#

My controlnet pre-processors look like this

image.png
upbeat crane
#

nothing in advanced controlnet huh

viral fox
#

so I have a few, just not the ComfyUi Preprocessor or the Apply controlnet dialouge ones

upbeat crane
#

yeah sorry wish i could be of better help, but im getting my ass handed to me by controlnets myself today

#

we both need a WIN bad right now

viral fox
#

nope nothing

upbeat crane
#

might just be devs but worth a shot

#

i love troubleshooting but comfy's learning curve is a little too steep for my patience level i think

#

even when i watch ballers on yt who are trying to give tutorials

#

half the time look like they have no idea what they are doing

#

Scott Detweiler is the dude tho, his channel is killer and he actually responds to messages quite a bit

#

on YT

#

works at stability and is comfy master

viral fox
#

Fiddlesticks I have lost my menu and manager popup for controlnet

upbeat crane
#

i bet if you tweeted or sent him a message he would tell u exactly what u needed if you cant find it elsewhere

#

enjoy digging through all those strands of spaghetti tryin to find it

#

seriously he's a GREAT dude

#

at the least dig through his yt channel cuz im sure he covers what ur tryin to do

#

there are 3 different ones here, another one you could try searching for is called Marigold

#

also i dont think diffusers CNs are compatible in comfy. you need to get the MIDAS pre-processor and use it with the T2iadapter_depth CN

#

illyasviel said to try and install Forge via StabilityMatrix for the cn issue WOW this is really nice

#

very cool

old parrot
#

Stablility matrix seems useless if you're only using Forge

upbeat crane
#

gunna play around with it, it looks interesting to me

#

im having an issue with the integrated controlnets in forge and the dev said to try it out

#

and if i can get forge working correctly it solves all my deforum issues with mult sdxl models and mult controlnets

#

basically im runnin around like a chicken with my head cut off tryin to solve issues lol

upbeat crane
#

okay yeah well that didn't work either, same issue....deleted

dusk minnow
#

Quick question, does anyone have a tutorial on how to install it properly for the 4090? I have the same installation from when i had a 3070, and i get 10 it/s with a 4090 wich seems a bit low

upbeat crane
#

10 sounds low wow

#

what commandline_args u using?

sweet basin
#

can someone post what to use for "set COMMANDLINE_ARGS=" nvidia 1080ti

dusk minnow
upbeat crane
#

--xformers only

#

thats fair, there isn't going to be any different type of installation or anything

#

search 4090 in the search box in this channel and compare your performance with other peoples

#

if you are pushing 10/its with sdxl models, u are doin good i think

#

seen a lot of people with a lot less

#

just gotta dial in your commandline args, which is most like --opt-sdp-attention --xformers

#

but not 100% might hit up CS1o when he gets back

woven canopy
upbeat crane
#

btw you dont need --xformers with the --opt. the --xformers only was meant for @sweet basin

#

i wonder what card is the cut off to where --xformers is antiquated

river flame
#

Hi! How do I install DWPose?

image.png image.png
karmic crown
#

can someone please explain where some people found info that they need to add the --autolaunch parameter?! I am honestly fascinated. a1111 will open a new tab in your browser automatically anyway, except when you use --listen or --share. those are the only cases in which you need to add --autolaunch 🤷🏻‍♂️

upbeat crane
#

@river flame time-stamped for your viewing pleasure: https://youtu.be/30b2k1p2CiE?t=168

YouTube
Olivio Sarikas
DW Pose for A1111 - Better than OpenPose ControlNET

DW Pose is much better than Open Pose Full. Especially the Hand Tracking works really well with DW Pose. This is a full review. Full Install Guide for DW Pose in A1111 for Stable Diffusion

links from the Video

DW Pose: https://github.com/IDEA-Research/DWPose
A1111 1.6 Guide https://youtu.be/1Ldc5kZKKFw?si=Hgb0rtVEhLxEQkkL
IMPORTANT: F...

▶ Play video
queen plank
#

Hey, I'm trying to run stable cascade on comfy, but I keep on getting this error

#

Error occurred when executing DiffusersStableCascade:

module diffusers.pipelines.stable_cascade has no attribute StableCascadeUnet

File "D:\ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\ComfyUI\execution.py", line 152, in recursive_execute
output_data, output_ui = get_output_data(obj, input_data_all)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\ComfyUI\execution.py", line 82, in get_output_data
return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\ComfyUI\execution.py", line 75, in map_node_over_list
results.append(getattr(obj, func)(**slice_dict(input_data_all, i)))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-DiffusersStableCascade\nodes.py", line 43, in process
self.prior = StableCascadePriorPipeline.from_pretrained("stabilityai/stable-cascade-prior", torch_dtype=torch.bfloat16).to(device)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\python_embeded\Lib\site-packages\huggingface_hub\utils_validators.py", line 118, in _inner_fn
return fn(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\src\diffusers\src\diffusers\pipelines\pipeline_utils.py", line 1265, in from_pretrained
loaded_sub_model = load_sub_model(
^^^^^^^^^^^^^^^
File "D:\ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\src\diffusers\src\diffusers\pipelines\pipeline_utils.py", line 444, in load_sub_model
class_obj, class_candidates = get_class_obj_and_candidates(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\src\diffusers\src\diffusers\pipelines\pipeline_utils.py", line 337, in get_class_obj_and_candidates
class_obj = getattr(pipeline_module, class_name)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "D:\ComfyUI_windows_portable_nvidia_cu121_or_cpu\ComfyUI_windows_portable\src\diffusers\src\diffusers\utils\import_utils.py", line 697, in getattr
raise AttributeError(f"module {self.name} has no attribute {name}")

river flame
upbeat crane
#

might be worth a shot, i dont use comfy

#

makes me wonder if it has an issue with the portable version

jaunty fulcrum
upbeat crane
#

@queen plank you may need to update your diffusers with pip install --upgrade diffusers

queen plank
upbeat crane
#

be an easy fix if it works

#

well NO

karmic crown
#

is there a deforum discord?

upbeat crane
queen plank
upbeat crane
#

just reading several pages that all having issues with comfy, cascade diffusers and see several people mentioning it fixed

#

but it is no guarantee obviously

queen plank
#

Alright, I did it, now time to test

upbeat crane
#

or do more research into diffusers version, searching comfyui PORTABLE and cascade, see what else you come up with first if you like

queen plank
#

Error occurred when executing DiffusersStableCascade:

module diffusers.pipelines.stable_cascade has no attribute StableCascadeUnet

hollow quartz
#

Hi, I am using Stable Diffusion in forge with a nvidia gtx 1080 graphics card. Whenever I try to press "Interrogate CLIP" an error pops up. Has anybody got an idea how I can get Interrogate CLIP to work for me? Do I need to have a better graphics card?

queen plank
upbeat crane
#

i bet it has something to do with a limitation of the portable version you are running

#

thats what i would focus on with research

queen plank
#

I'll dig into it

upbeat crane
#

just a thought

karmic crown
hollow quartz
karmic crown
#

you probably do not have enough vram

upbeat crane
#

i dont think i've ever used that in my life

karmic crown
#

this will tell it to use cpu instead

queen plank
upbeat crane
#

i ended up finding a cool little standalone tool called SD Prompt Reader and ended up kinda digging it

#

althogh interrogating clip obvi makes more sense

river flame
upbeat crane
#

possible a VAE or lack of one?

river flame
upbeat crane
#

im always havin VAEs mess with stuff and takes me a min to figure it out, just something to consider

#

im not real sure

#

if vae is on AUTO try setting to none maybe?

#

was a long-shot

#

never worked with that before but i hear good things

#

can you try a diff checkpoint maybe

river flame
river flame
upbeat crane
#

i'd like to hear what fixes is when/if you do tho, good luck man

river flame
#

oh no it worked from the 2nd try

image.png
upbeat crane
#

nice

#

thats a strange lookin broad

hollow quartz
ornate elk
karmic crown
sweet basin
#

yeah thx that what I told to use

upbeat crane
#

what does the error say @hollow quartz

ornate elk
river flame
karmic crown
#

@hollow quartz probably those "--xformers --medvram-sdxl --no-half-vae --theme dark --use-cpu interrogate"

#

what was your gpu again? CS1o know those better than me

#

i know what you need for mac 🙂

ornate elk
hollow quartz
# upbeat crane what does the error say <@1158769414611673134>

I don't really know how to exactly trace errors lol, this is what appears in cmd:
Traceback (most recent call last):
File "I:\forge\stable-diffusion-webui-forge\venv\lib\site-packages\gradio\routes.py", line 488, in run_predict
output = await app.get_blocks().process_api(
File "I:\forge\stable-diffusion-webui-forge\venv\lib\site-packages\gradio\blocks.py", line 1431, in process_api
result = await self.call_function(
File "I:\forge\stable-diffusion-webui-forge\venv\lib\site-packages\gradio\blocks.py", line 1103, in call_function
prediction = await anyio.to_thread.run_sync(
File "I:\forge\stable-diffusion-webui-forge\venv\lib\site-packages\anyio\to_thread.py", line 33, in run_sync
return await get_asynclib().run_sync_in_worker_thread(
File "I:\forge\stable-diffusion-webui-forge\venv\lib\site-packages\anyio_backends_asyncio.py", line 877, in run_sync_in_worker_thread
return await future
File "I:\forge\stable-diffusion-webui-forge\venv\lib\site-packages\anyio_backends_asyncio.py", line 807, in run
result = context.run(func, *args)
File "I:\forge\stable-diffusion-webui-forge\venv\lib\site-packages\gradio\utils.py", line 707, in wrapper
response = f(*args, **kwargs)
File "I:\forge\stable-diffusion-webui-forge\modules\ui.py", line 837, in <lambda>
fn=lambda *args: process_interrogate(interrogate, *args),
File "I:\forge\stable-diffusion-webui-forge\modules\ui.py", line 113, in process_interrogate
return [interrogation_function(ii_singles[mode]), None]
File "I:\forge\stable-diffusion-webui-forge\modules\ui.py", line 135, in interrogate
prompt = shared.interrogator.interrogate(image.convert("RGB"))
AttributeError: 'NoneType' object has no attribute 'convert'

river flame
ornate elk
river flame
upbeat crane
#

could a themes be interfering by chance @hollow quartz ?

hollow quartz
hollow quartz
upbeat crane
#

like a skin or something for the user interface

#

could also be an extension, you can sometimes uncheck an extension or two and reload and test if that may be a possibility

hollow quartz
upbeat crane
#

i finally solved me dang Forge issue not populating the built in control nets!! yay

#

was a strange combo of things

#

whats weird for me now, is that i get error messages FOR actually disabling a theme

#

Exception in ASGI application
Traceback (most recent call last):
File "D:\AI\stable-diffusion-webui-forge\stable-diffusion-webui-forge\venv\lib\site-packages\uvicorn\protocols\http\h11_impl.py", line 412, in run_asgi
result = await app( # type: ignore[func-returns-value]
File "D:\AI\stable-diffusion-webui-forge\stable-diffusion-webui-forge\venv\lib\site-packages\uvicorn\middleware\proxy_headers.py", line 84, in call
return await self.app(scope, receive, send)
File "D:\AI\stable-diffusion-webui-forge\stable-diffusion-webui-forge\venv\lib\site-packages\fastapi\applications.py", line 273, in call
await super().call(scope, receive, send)

#

and it goes on and on ....

#

was an extention that lets me change the default web browser the UI opens in

#

i never used it, or made any changes to it's settings, but when i disable it, i get these error messages

#

sorry not disabling a theme, i meant DISABLING AN EXTENTION

#

causes the errror

#

sd-webui-custom-autolaunch is the name of it, really obscure and never even used it

#

unchecked it, even tried removing the folder, then the ui wouldnt load

#

so i restored the folder, with it unchecked, and the UI loads but im stuck with this

cobalt ingot
#

Does anyone have experience using Flowframes? When I try to interpolate my photos to make a video, not only does it play the image out of order, but some of them don't smoothly change to the next frame and it's cutting out half of my photos and not including them at all...

#

I have seen a few things on Google about playing them out of order, but nothing I Google says anything about missing half the photos

karmic crown
ornate elk
ornate elk
#

From time to time

upbeat crane
#

will do ty

cobalt ingot
# ornate elk Hey do you have the issue that you see the first image in the video flickering i...

It doesn't even start with the image that I want it to start with. It shows it for a second and then switches to another random frame. It doesn't merge with the next frame like it should. It does merger the frames smoothly in later parts of the video, but not for them all and I think it's because it's playing them out of order and cutting out a bunch of photos that are necessary for it to flow

#

It's like it plays a third of the photos in order like it's supposed and the rest are random or simply not included

upbeat crane
#

not 100% sure how i fixed it, but i did a few things....deleted venv and let it rebuild and noticed some messages about certain things needing updating, so i ended up doing a pip upgrade and a pip upgrade to FastAPI.. the next time i loaded the UI, i was too stoked when i clicked on the Pre-Procecssor drop down in Deforum and this HUGE list popped up. could hardly believe it. still haven't gotten around to re-running my project yet using Canny and Depth, but i have a real strong feeling that its gunna work now. i have four different versions of the top 3-4 CNs now @ornate elk

#

get your hands on Topaz Video AI @cobalt ingot

ornate elk
#

but if you input a video it will work

reef smelt
#

should stable diffusion use a lot of ram?

#

like 10gb+?

upbeat crane
#

when is rendering or idle? that sounds pretty high

reef smelt
#

well it was at 20gb before on idle but after rendering

#

almost crashed pc 😭

upbeat crane
#

you could monitor your system resources and uncheck extensions maybe just for testing

reef smelt
#

hmm

#

I haven't added anything I think...

#

weird...

ornate elk
reef smelt
#

I see

upbeat crane
#

exactly

reef smelt
#

then I understand

upbeat crane
#

i came to expect that behavior when using sdxl models, controlnets in combo were out of the question

ornate elk
upbeat crane
#

so im real happy to have forge back working

reef smelt
#

does forge have drawbacks?

upbeat crane
#

install Web ForgeUI @reef smelt it is WAYYY better on resources, night and day

#

WebUI Forge sorry

reef smelt
#

is it faster?

upbeat crane
#

its still A1111 just better optimized and yeah its faster

cobalt ingot
upbeat crane
#

depends on the card, look into it, search it on YT listen to what ppl say

reef smelt
upbeat crane
#

has some issues, which i have been working through, but with SDXL, i just couldnt continue like that so im moving forward only with Forge and more comfyui here soon

woven canopy
upbeat crane
#

u should def give it a shot tho, takes 3 min to install, worth a try no doubt

woven canopy
#

Some of these tokens though LMAO....

Screenshot_20240307_120027_Chrome.png
upbeat crane
#

punchable expression hahaa

ornate elk
# cobalt ingot I am using the .40 version

oh okay, then idk, i also tried Enhancr as it has also interpolate option. But not for batch input.
Lastly i ended up doing interpolation manually with rife-ncc-vulkan and ffmpeg.
But would also recommend checking out Waifu4x-extension-gui

upbeat crane
#

thats good

#

someone mentioned that i didnt know ffmpeg did interpolating till today

ornate elk
woven canopy
#

Oh, I didn't have no half vae, I'll try that next

cobalt ingot
hollow quartz
# upbeat crane could also be an extension, you can sometimes uncheck an extension or two and re...

Tried disabling many extensions individually, tried disabling all extensions, but sadly nothing worked and I still got the same error. This is the error if anyone is wondering:
Traceback (most recent call last):
File "I:\forge\stable-diffusion-webui-forge\venv\lib\site-packages\gradio\routes.py", line 488, in run_predict
output = await app.get_blocks().process_api(
File "I:\forge\stable-diffusion-webui-forge\venv\lib\site-packages\gradio\blocks.py", line 1431, in process_api
result = await self.call_function(
File "I:\forge\stable-diffusion-webui-forge\venv\lib\site-packages\gradio\blocks.py", line 1103, in call_function
prediction = await anyio.to_thread.run_sync(
File "I:\forge\stable-diffusion-webui-forge\venv\lib\site-packages\anyio\to_thread.py", line 33, in run_sync
return await get_asynclib().run_sync_in_worker_thread(
File "I:\forge\stable-diffusion-webui-forge\venv\lib\site-packages\anyio_backends_asyncio.py", line 877, in run_sync_in_worker_thread
return await future
File "I:\forge\stable-diffusion-webui-forge\venv\lib\site-packages\anyio_backends_asyncio.py", line 807, in run
result = context.run(func, *args)
File "I:\forge\stable-diffusion-webui-forge\venv\lib\site-packages\gradio\utils.py", line 707, in wrapper
response = f(*args, **kwargs)
File "I:\forge\stable-diffusion-webui-forge\modules\ui.py", line 837, in <lambda>
fn=lambda *args: process_interrogate(interrogate, *args),
File "I:\forge\stable-diffusion-webui-forge\modules\ui.py", line 113, in process_interrogate
return [interrogation_function(ii_singles[mode]), None]
File "I:\forge\stable-diffusion-webui-forge\modules\ui.py", line 135, in interrogate
prompt = shared.interrogator.interrogate(image.convert("RGB"))
AttributeError: 'NoneType' object has no attribute 'convert'

upbeat crane
#

you get that just for clicking the INTERROGATE CLIP button?

#

but you could try deleting the VENV folder in your stable diffusion webui folder

#

then the next time you launch it will rebuilt itself

#

or do a reinstall maybe

#

i use this standablone EXE, i love this little prompt reader tool, you just

#

not sure if it does exactly the same thing, i never really used what your talking about

night orchid
#

Have any of you tried Reactor and mov2mov?

#

when I click generate, notting happen

reef smelt
#

how is this even possible? it's a compressed archive-

image.png
hollow quartz
reef smelt
#

I think I got it now

#

but tell

upbeat crane
#

its just a tool that i keep open in one window, have explorer showing all my images and i just drag any photo into it and shows me all the meta data, has copy tools etc..... your probably looking for like something that might NOT have any meta data and have the AI actually analze the clip and tell you about it, i suppose

#

apologizes if it didnt apply, i need to give clip interrogator a try later today

reef smelt
#

so.. if it has data it will show that data but if it doesn't it'll guess?

upbeat crane
#

maybe there are like OTHER tools you could use, maybe google stable diffusion clip interrogation tools, or sarch through your extentions tab. surely something else out there wil work for ya

#

correct

reef smelt
#

😭 your confusing me

upbeat crane
#

thought you said, if it doesnt ....like no it doesnt show anything

#

no it doesnt guess

reef smelt
#

I'm not looking for features lol

upbeat crane
#

read my above comment then, search for an alternate tool maybe you can find a diffferent way of accomplishing the task

reef smelt
#

I'm not sure.. why I would do that?

#

I was just wondering lol

upbeat crane
#

unless its completely jammin up your stable diffusion, and if so, i recooment removing the VENV folder, or doing full reinstall

#

i just obviously didnt fully understand what clip interrogation was, my bad

reef smelt
#

😔 I swear like your talking to someone else

upbeat crane
#

i was only trying to help you, and i didnt fully understand what clip interrogation was, because i had never used it before, and i thought maybe this tool i used would be of assistance to you, but it is not, so i gave you several other suggestions that might help you with your problem. cheers have a good day

reef smelt
#

Wut- I didn't have.. a problem from the start?

#

think you mixed up me with someone else 👀

upbeat crane
#

yeah i think so, apologies

reef smelt
#

it's okay I just find it funny

upbeat crane
#

me too

reef smelt
#

what's vae-approx

upbeat crane
#

it's a dollar tree version of a variational autoencoder

reef smelt
#

not sure what that means

upbeat crane
#

vae's to me seem like they help a lot with color and contrast and kinda make images pop a bit better

#

im no expert on it, just kinda what ive seen, you can load them along with a checkpoint/model

woven canopy
upbeat crane
#

directly to the right or below your model loader

#

heres a bunch of nerd data on em https://en.wikipedia.org/wiki/Variational_autoencoder

Variational autoencoder

In machine learning, a variational autoencoder (VAE) is an artificial neural network architecture introduced by Diederik P. Kingma and Max Welling. It is part of the families of probabilistic graphical models and variational Bayesian methods.Variational autoencoders are often associated with the autoencoder model because of its architectural aff...

reef smelt
ornate elk
reef smelt
#

are any of these better than xformers?

image.png
karmic crown
#

if you are using nvidia, probably not

reef smelt
#

I am using nvidia

upbeat crane
#

--opt-sdp-attention but depends on the card

reef smelt
#

rtx 2060

#

not super

#

it don't got the triangle s on it

#

does forge always use xformers or?

upbeat crane
#

you should use --xformers --med-vram

reef smelt
#

well.. I don't think I can use -med-vram

#

anymore at least

kind sapphire
upbeat crane
#

what program/tool are you using Rem?

#

a1111? forge?

#

comfy

reef smelt
#

I'm using forge now

upbeat crane
#

yeah how you liking it so far?

#

how does it run for ya

reef smelt
#

haven't used it actually-

#

don't know if it's actually working as it should...

#

quite concerning

upbeat crane
#

hmmm

reef smelt
#

Traceback (most recent call last):
File "C:\Users\rem\Desktop\a\forge\system\python\lib\site-packages\gradio\routes.py", line 488, in run_predict
output = await app.get_blocks().process_api(
File "C:\Users\rem\Desktop\a\forge\system\python\lib\site-packages\gradio\blocks.py", line 1429, in process_api
inputs = self.preprocess_data(fn_index, inputs, state)
File "C:\Users\rem\Desktop\a\forge\system\python\lib\site-packages\gradio\blocks.py", line 1222, in preprocess_data
self.validate_inputs(fn_index, inputs)
File "C:\Users\rem\Desktop\a\forge\system\python\lib\site-packages\gradio\blocks.py", line 1209, in validate_inputs
raise ValueError(
ValueError: An event handler (ControlNetUnit) didn't receive enough input values (needed: 22, got: 1).

#

duno what that's supposed ot mean but ok

#

yea it's unresponsive...

#

hmm

upbeat crane
#

yeah im getting traceback errors too in forge

reef smelt
#

strange.. but sure

upbeat crane
#

its 100$ for sure an issue with an extention i had

reef smelt
#

what?

upbeat crane
#

even with it it uncheck, its done something i cant quite figure out

#

im working with deforum right now in forge

#

and been troubleshooting controlnets all day

#

the only thing i can say right now, is im 99$ sure that diffuser CNs dont work in forge

#

im starting to test the other ones, but i had a major issue with the built in one not loading in to the drop downs

#

i finally loved that, so bout to know something here soon i hope

night orchid
#

@upbeat crane you dont have a alongside installation of "normal" SD ?

upbeat crane
#

guess what im trying to say is, i ALSO am not sure that Forge is working the way it should either

#

well, i did, actually yeah i still do

#

like calling to the A1111 and sharing checkpoints and controlnets?

#

you think i should cut all that off and isolate it?

night orchid
#

no thats a shame but you cant

night orchid
upbeat crane
#

yeah i spent a day and half troubleshooting the built in CNs not appearing

#

i finally fixed that, now i see all the ones i should

#

but my first deforum render, got the same controlnet error

#

gunna test a bit more, then im pretty close to giving up on using deforum with forge, until its been out a few more months and bugs get sorted

night orchid
#

Forge is so good with loading of models

upbeat crane
#

why did you ask me "you dont have a alongside installation of "normal" SD" ?

#

oh you meant, like as another option?

night orchid
#

because if forge is not working you could test there

upbeat crane
#

oh i see, i thought you had some troubleshooting hint or something

karmic crown
night orchid
#

you can? wow thats nice

upbeat crane
#

cuz i use Deforum, make hq animations that i upscale to 4k, and so i have to use XL checkpoints mainly

#

and using sdxl, trying to load XL checkpoints, multiple controlnets etc, just FLOPS on resources to the point

#

that I had to move on to something else, forge seemed promising AF at first

karmic crown
upbeat crane
#

now the illusion is wearing off. its AMAZING for IMAGE rendering and memory optimization

#

so ill def be using Forge for that moving forward

reef smelt
upbeat crane
#

%

karmic crown
#

100€ 🙂

reef smelt
#

oh yea I moved all my extensions and must have forgotten to tell which should be enabled

#

think that's it

upbeat crane
#

going to go poke around deforum discord later tonight to see if i can find one person who says they are using multiple controlnets with deforum inside of Forge WebUI. i would like to meet them

#

yup dang i think im done trying on this, nothing else to try

#

*** Error running process_before_every_sampling: D:\AI\stable-diffusion-webui-forge\stable-diffusion-webui-forge\extensions-builtin\sd_forge_controlnet\scripts\controlnet.py
Traceback (most recent call last):
File "D:\AI\stable-diffusion-webui-forge\stable-diffusion-webui-forge\modules\scripts.py", line 835, in process_before_every_sampling
script.process_before_every_sampling(p, *script_args, **kwargs)

#

thats using the BUILT-IN controlnets too ugh

#

i also do realize that Deforum is experimental in forge, but ive done enough experimenting lol

karmic crown
upbeat crane
#

yeah

#

Forge is really new

#

they will get it all lined out with deforum at some point i bet

karmic crown
#

i like it, but it works slower than a1111 on my mac 🙈

upbeat crane
#

oh really?

karmic crown
#

a1111 2it/s, forge and comfy 1.5it/s 🙈

upbeat crane
#

i have a 3060 with 12gb of ram, and using Forge for images with control nets is WAY better than A1111 for me

#

ill go back to A1111 im just gunna have to let go of rendering my deforum animations using XL checkpoints, no biggie

karmic crown
#

but i just tested 1.5 default model so far on this new mac

upbeat crane
#

all the 1.5 models work pretty well and the controlnets without bogging the program down with OOMs

#

ill survive but the XL models are just SOOOO beautiful with deforum, like jaw dropping

reef smelt
#

well the efficency isn't noticeable on higher end pcs

upbeat crane
#

i have topaz video so i can do some nice upscaling and i didnt have this back when i was working with 1.5 models

reef smelt
#

works nicely

upbeat crane
#

so, ill be fine, just gotta bide my time for a bit and let things unfold

karmic crown
upbeat crane
#

we push the limits pretty hard lol

#

oh Deforum doesnt work with 1.8.1

#

it vanishes

#

so all the deforum nerds have had to roll back to 1.7

karmic crown
#

1.8.1 does not exist

#

it works

#

we fixed that

#

i helped

upbeat crane
#

1.8 whatever its called

reef smelt
karmic crown
upbeat crane
#

i like Sarge's workflows in Comfui, i just got those installed, looks pretty nice

#

i think i could work with his setup

karmic crown
#

deforum now works on 1.8.0, but does not work on older versions

upbeat crane
#

oh nice

karmic crown
upbeat crane
#

im not even sure what alls new from 1.7 to 1.8, hopefully good things

#

it aint gunna take much, cuz a1111 1.7and windows 11 have issues

reef smelt
karmic crown
# reef smelt what resolution?

as i said to CS1o, who need a PC when you can play CS2 with 120 FPS on your Mac with 3 layers of emulation on 1512x982 120Hz with max details 🤣

reef smelt
#

CS1o is in the past- I mean.. Nostalgia is the best

upbeat crane
#

i just sold a ton of 2014 skins and made 160 bucks

#

i got pretty thick into crate gamblin back in the day

#

hehehe

reef smelt
#

😔 tech-support is now just csgo

upbeat crane
#

this was me lookin to score crates

karmic crown
#

my son is one of lucky kids who can say "my mom and dad played CS, as well as me" 🙂

reef smelt
#

aww..

upbeat crane
#

a family that games together stays together

#

thats lovely

#

whatcho been playin lately viking?

#

im playin Forza5, Starfield, HarryPotter and the new Ratchet and Clank

karmic crown
upbeat crane
#

nice, appreciate it

karmic crown
#

that will happen if you try to use latest deforum on 1.7.0

upbeat crane
#

ohh so just DONT update my extentions then

#

i havent launched A111 since 1.8 came out

karmic crown
#

just update both

#

and it will work

upbeat crane
#

will do

#

im tired of troubleshooting time to get back to creating

#

there are some amazing 1.5 models to work with

#

just gotta make em count

reef smelt
#

what sampling method is best?

karmic crown
#

switch to mac and use my script, you can get working version in a few minutes 🙈

karmic crown
upbeat crane
#

ive really been like DMP++ 3M SDE with XL models, but my main is probably ++2m karras

karmic crown
#

just follow the recommendation from the model author about sampler and steps

upbeat crane
#

right, some models work better with certain samplers, some dont work well with really GREAT LORAs

#

gotta be flexible and a big-time tinkerer

noble belfry
#

sampling reminds me of people who buy gold audio cables and swear they can hear the difference

upbeat crane
#

haha

#

nice

#

Bro. check this out MY CABLES, right?: 10karats gold

#

AFRICAN gold bruh

reef smelt
noble belfry
#

i mean ok, sure theres a difference between a piece of string and a cup audio quality and an actual cable, but beyond that i think a lot is percieved difference

noble belfry
#

jk

reef smelt
#

Lol

karmic crown
#

blah

upbeat crane
#

the biggest thing with samplers with me, is that some are way better at following prompting

#

but they are pretty random as to like THIS one doesnt work with that LORA or this one

#

its all maddening at times

karmic crown
#

add LCM to mix and you can just cry in some cases, when cfg is 1.5

upbeat crane
#

it all does feel like a gigantic boul of soupy randomness at times

#

i alwasy think, like i have BAD days with it

#

then all the sudden, i will be DIALED the F in, its wild

karmic crown
#

btw. this is tech-support, not general-chat
just saying 🙂
we should stick to the topic here

upbeat crane
#

quit bein personable n shit

#

lol

karmic crown
#

i refuse to isntall basicsr

#

a1111 removed that, but defourm still use it, so it will not work on clean 1.8.0 install
i made a pr, but no one even reacting to that 🙈

upbeat crane
#

i like euler a at like 16 steps in deforum a lot

#

like a sweet spot for me

karmic crown
#

oh, well, a1111-setup v0.0.7 needs to be made i guess

upbeat crane
#

have you ever used Parseq before?

#

you can match music up to the frame generations and interpolate with it

#

and it helps with the camera and canvas controls by just using a mouse and moving functions on a wave form

#

started playin with that recently pretty nice tool

glad fiber
#

so, i am trying to train a lora with 11 images, but when the training ends the output images are nothing like i want. The avr_loss was around 0.1 and i was wondering if it has something to do with the images not working

#

now i am trying to train with 1e-6 learning rate and the avr_loss is aroung 0.09

#

i am on a gtx 1650 4gb, so i cant increase a lot of stuff here

upbeat crane
#

are you using a specific tool or anything to do it?

karmic crown
#

i cannot find the guide that i saw that recommends 30 images if i remember correctly, and there was some explaination about amount of images in each group (like portrait, body shots...)

upbeat crane
#

i dont know anything about that subject unfortunately

karmic crown
#

for examole 10 of those, 4 of those...

upbeat crane
#

just wondering if you were using somethign specific sometimes whoever made the tool will have a discord

#

i usually just start by watching every avail yt tutorial that may exist on a subject

#

are you using Kohya ? err

glad fiber
upbeat crane
#

i know there is a bunch of YT tutorials on that

glad fiber
#

i have followed a lot of them

upbeat crane
#

that does sound like an awfully small data sample too but not sure man, wish u the best of luck

#

keep asking around somebody will know something hopefully

glad fiber
#

but like, even with a small dataset, it should at least give me a glimpse of what i want

quick forge
#

ok....so ForgeUI is genuinely kind of pissing me off a little at this point ....

upbeat crane
#

yeah thats been the discussion, i was so excited about it, but after these last two days

#

im pretty much giving up on it for what i need it for

#

its BUG city

quick forge
#

because it's doing the stupidest stuff

upbeat crane
#

yeah i cant get it to work with controlnets for the life of me

#

in deforum

#

fix one things then two more break

#

it's aggrivating me too you aren't alone

#

it was so promising too and seems like it runs so smooth at first

#

till you start using it

quick forge
#

ForgeUI?

upbeat crane
#

yeah

quick forge
#

if we're talking about the same thing, my problem is that it just keeps ignoring any settings changes I make, and then when it does change them, I don't seem to be able to revert back

night orchid
#

@upbeat crane what is the different between Forge and the normal 1111 ?

quick forge
#

now when I run it, it's cloning itself into the "stable-diffusion-webui" sub folder within it's install folder, and then just running A1111 from that folder

karmic crown
#

@upbeat crane see this

quick forge
#

and is faster

upbeat crane
#

forge still based off the A1111 interface, its really good at optimizing memory, its really FAST n such but its buggy as hell

night orchid
#

the image shows SD 3

quick forge
#

I deleted and re-cloned the repo, but it's still cloning it-fucking-self even though it's a new install

upbeat crane
#

i needed it to push XL models with multiple controlnets doing animation cuz a1111 is terrible at it for me

#

ive reinstalled it five times in 72 hours

karmic crown
#

i reisntalled 5 times in last 5 minutes 🙂

#

but i am developing a new version of install script 🙂

night orchid
#

I was trying to use forge with mov2mov, not working

night orchid
quick forge
#

All I wanted to do is run it along side other Ui's I have installed and have it share the venv and models/extensions/etc folders like the others do.

#

but apparently you have to be some kind of fucking wizard to make it work

#

and believe me, I'm no inexperienced user with computers and this sort of thing

karmic crown
#

if you are talking about forge it is litterary 1 command line param

upbeat crane
#

preach it my man im with you

#

im giving up on it

quick forge
upbeat crane
#

i will still use it to do image generation with XL models tho

karmic crown
#

to use everything from a1111 in forge

upbeat crane
#

but Forge is like ALLERGIC to controlnets from my perspective

karmic crown
#

models, loras...

night orchid
quick forge
upbeat crane
#

just wipe that VENV folder, it will put it back right

quick forge
#

even though CLONEDIR is commented out

quick forge
#

still cloning A111 and not actually running "forge"

#

all I did was uncomment that clonedir line once, and now it won't stop doing it

upbeat crane
#

the only thing i actually havent done

#

is to move over the checkpoints and CNs directly to Forge, and like isolate it from linking to A1111 then testing

karmic crown
quick forge
#

I'm running on Ubuntu btw

upbeat crane
#

lol

#

when i FIRST installed it, i ran a perfect deforum animation

#

used four loras, canny and depth controlnets, did checkpoint scheduling

#

ran perfectly

#

and since then, havent gotten it to work again

#

makes zero sense

#

so apparantly it is possible, im just not smart enough to work it

night orchid
#

[ERROR:0@0.192] global cap_ffmpeg_impl.hpp:1268 open Could not find decoder for codec_id=61
[ERROR:0@0.194] global cap_ffmpeg_impl.hpp:1317 open VIDEOIO/FFMPEG: Failed to initialize VideoCapture
[ERROR:0@0.344] global cap_ffmpeg_impl.hpp:1268 open Could not find decoder for codec_id=61
[ERROR:0@0.344] global cap_ffmpeg_impl.hpp:1317 open VIDEOIO/FFMPEG: Failed to initialize VideoCapture
[ERROR:0@0.744] global cap_ffmpeg_impl.hpp:1268 open Could not find decoder for codec_id=61
[ERROR:0@0.744] global cap_ffmpeg_impl.hpp:1317 open VIDEOIO/FFMPEG: Failed to initialize VideoCapture

#

whats does this means?

upbeat crane
#

probably means you dont have FFMPEG placed properly to system path, so that whatever tool that is can call to it

#

system environment variables, google it, check and see that ffmpeg.exe is placed into system path

#

you should be able to open up a terminal and just type ffmpeg and it see it

#

if not, you need to add it to path

night orchid
#

isn't that correct?

upbeat crane
#

click down below where it says PATH

#

and add it there as well

vocal burrow
night orchid
#

I just follow a tutorial

quick forge
#

like how the actual fuck do I stop this:

Clone stable-diffusion-webui
################################################################
Cloning into 'stable-diffusion-webui'...
remote: Enumerating objects: 31936, done.
remote: Counting objects: 100% (405/405), done.
remote: Compressing objects: 100% (237/237), done.
remote: Total 31936 (delta 227), reused 283 (delta 156), pack-reused 31531
Receiving objects: 100% (31936/31936), 34.17 MiB | 41.90 MiB/s, done.
Resolving deltas: 100% (22335/22335), done.

upbeat crane
#

yeah it tells you to add it to path, and you see path and added it

#

but you gotta do that one there on the bottom too, if you havent

#

hopefully that will fix it

night orchid
#

@upbeat crane its greyd out

upbeat crane
#

close your window on the right?

night orchid
upbeat crane
#

if that doesnt help you may need to set your user some admin level priveledges maybe

night orchid
#

hmm ok thanks

upbeat crane
night orchid
#

it okay now

upbeat crane
#

I was indeed accessing the env vars via the Start menu 🤦‍♂️. The other option, "Edit the system environment variables" will open a different menu -- just click "Environment variables..." at the bottom. (Thank you for pointing this out ⭐)

#

ok cool

karmic crown
upbeat crane
#

wow, i actually have a deforum animation rendering rn on an xl model, it has the controlnet errors tho, but still NICE

#

in forge

#

it just works when it feels like

#

AI is messing with my mind

night orchid
#

its so frustrating, when I run mov2mov and Reactor I get notting, notting happen, no error

upbeat crane
#

yeah its actually rendering nicely for the first time today rn

#

but still showing this "Neither the base nor the masking frames for ControlNet were found. Using the regular pipeline
2024-03-07 16:41:45,326 - ControlNet - INFO - ControlNet Input Mode: InputMode.SIMPLE"

quick forge
upbeat crane
#

ive tried four different canny's and four different depths

karmic crown
quick forge
#

not bat ... sh

karmic crown
#

ok

#

same thing, diffrent os 🙂

night orchid
#

hmm I had Controlnet Ext. enabled

quick forge
#

the whole thing is commented out.....

#

Install directory without trailing slash

#install_dir="/home/$(whoami)"

Name of the subdirectory

#clone_dir="stable-diffusion-webui"

Commandline arguments for webui.py, for example: export COMMANDLINE_ARGS="--medvram --opt-split-attention"

#export COMMANDLINE_ARGS=""

python3 executable

#python_cmd="python3"

git executable

#export GIT="git"

python3 venv without trailing slash (defaults to ${install_dir}/${clone_dir}/venv)

#venv_dir="../venv"

script to launch to start the app

#export LAUNCH_SCRIPT="launch.py"

install command for torch

#export TORCH_COMMAND="pip install torch==1.12.1+cu113 --extra-index-url https://download.pytorch.org/whl/cu113"

Requirements file to use for stable-diffusion-webui

#export REQS_FILE="requirements_versions.txt"

Fixed git repos

#export K_DIFFUSION_PACKAGE=""
#export GFPGAN_PACKAGE=""

Fixed git commits

#export STABLE_DIFFUSION_COMMIT_HASH=""
#export CODEFORMER_COMMIT_HASH=""
#export BLIP_COMMIT_HASH=""

Uncomment to enable accelerated launch

#export ACCELERATE="True"

Uncomment to disable TCMalloc

#export NO_TCMALLOC="True"

#

wtf?

#

that's weird lol

upbeat crane
#

you dont edit that file

#

right?

quick forge
#

Neat

#

#Install directory without trailing slash
#install_dir="/home/$(whoami)"

#Name of the subdirectory
#clone_dir="stable-diffusion-webui"

#Commandline arguments for webui.py, for example: export COMMANDLINE_ARGS="--medvram --opt-split-attention"
#export COMMANDLINE_ARGS=""

#python3 executable
#python_cmd="python3"

#git executable
#export GIT="git"

#python3 venv without trailing slash (defaults to ${install_dir}/${clone_dir}/venv)
#venv_dir="../venv"

#script to launch to start the app
#export LAUNCH_SCRIPT="launch.py"

#install command for torch
#export TORCH_COMMAND="pip install torch==1.12.1+cu113 --extra-index-url https://download.pytorch.org/whl/cu113"

#Requirements file to use for stable-diffusion-webui
#export REQS_FILE="requirements_versions.txt"

#Fixed git repos
#export K_DIFFUSION_PACKAGE=""
#export GFPGAN_PACKAGE=""

#Fixed git commits
#export STABLE_DIFFUSION_COMMIT_HASH=""
#export CODEFORMER_COMMIT_HASH=""
#export BLIP_COMMIT_HASH=""

#Uncomment to enable accelerated launch
#export ACCELERATE="True"

#Uncomment to disable TCMalloc
#export NO_TCMALLOC="True"

karmic crown
#

please format text with 3 backtics

upbeat crane
#

i must be lost here, so ill shut up

karmic crown
upbeat crane
#

thats not gunna do anything

karmic crown
#

3 of those before and after to format text properly, so i can read it

#

i cant see your code properly like this

upbeat crane
#

dang there i was saying nice things about Forge and it crashed during my animation

#

with all kinds of new errors i've never seen before hahaa

karmic crown
#

we are currently fixing deforum for a1111 to work on both 1.7.0 and 1.8.0, live on discord 🙈

quick forge
upbeat crane
#

oh makes sense

#

been chattin with some folks in desforum talking about all this

#

i imagine mov2mov same boat as deforum

quick forge
# 
#install_dir="/home/$(whoami)"

#Name of the subdirectory
#clone_dir="stable-diffusion-webui"

#Commandline arguments for webui.py, for example: export COMMANDLINE_ARGS="--medvram --opt-split-attention"
#export COMMANDLINE_ARGS=""

#python3 executable
#python_cmd="python3"

#git executable
#export GIT="git"

#python3 venv without trailing slash (defaults to ${install_dir}/${clone_dir}/venv)
#venv_dir="../venv"

#script to launch to start the app
#export LAUNCH_SCRIPT="launch.py"

#install command for torch
#export TORCH_COMMAND="pip install torch==1.12.1+cu113 --extra-index-url https://download.pytorch.org/whl/cu113"

#Requirements file to use for stable-diffusion-webui
#export REQS_FILE="requirements_versions.txt"

#Fixed git repos
#export K_DIFFUSION_PACKAGE=""
#export GFPGAN_PACKAGE=""

#Fixed git commits
#export STABLE_DIFFUSION_COMMIT_HASH=""
#export CODEFORMER_COMMIT_HASH=""
#export BLIP_COMMIT_HASH=""

#Uncomment to enable accelerated launch
#export ACCELERATE="True"

#Uncomment to disable TCMalloc
#export NO_TCMALLOC="True"```
#

oh neat

upbeat crane
#

since illyviasl is the dude who made forge and all those controlnets

quick forge
#

never knew that

upbeat crane
#

yeah

#

so, time to walk away from forge, come back at it in a few months

#

deforum for example, in a1111 ppl baically only run it with ALL ext disabled rn until further notice

#

so, im gunna do that, then try to push some 1.5 checkpoints and 1.5 controlnets

#

see if i can get A1111 back on track

#

he seems stressed lol

#

deforum has VIDEO OUTPUT btw you should jump in the discord

#

i meant INPUT

quick forge
#

what even is deforum?

onyx solar
#

deforum is where delakers used to play

#

🥁 🥁 💥

upbeat crane
#

here is a great representation of what deforum does

#

although it does a LOT of different things, this is one thing you can do

#

you can use all checkpoints/models etc to use Text2Video or apply diffusion to existing video input

#

thats a 4k video btw, crank it up a bit its default 360p

#

so its about as unlimited in theory as is image generation using SD

#

since you use all of the same Models and LORAS you just do it with video

#

TankerMedia is awesome btw some cool videos he does

quick forge
#

that's also cool af...

#

i just want forgeui to like....work though lol

upbeat crane
#

yeah if it worked like it can, it will be the best UI ever for deforum

#

they will get it there, for SURE

quick forge
#

It would be swell if it would just stop cloning itself/A1111 into a sub folder

upbeat crane
#

be a whole lot cooler if it did man

quick forge
upbeat crane
#

have a good night good chattin with ya, gunna go grab some food

karmic crown
#

i want to see what files have been changed

#

deforum will work on both 1.7.0 and 1.8.0

#

we are still fixing some minor issues

#

actually i am currently just help with testing

#

someone finally listened to me 🙂

upbeat crane
#

i believe it "works" using the default prompts, with 1.5 models and 0 controlnets

#

im back rendering now with it, and it wont even take 1.5 controlnets

#

please ping me if you can render with deforum with XL model and controlnet

#

i kinda think it's not worth doing anymore unless we can do those things

karmic crown
#

this is a1111 version, i have no idea if they will do the same with forge version or not

upbeat crane
#

yeah im back to a1111, i have given up completely on Forge

#

it's all just so strange because it wasnt like this 3 months ago

#

it's probably that there are so many people creating models now

#

and if they aren't trained in a very specific way, they break with deforum and controlnets

#

but thats just a wild guess, id just be happy to get back to where i used to be at this point

#

we'll get there just hit me up if you come up with anything interesting, and i shall do the same

#

man, one dude i follow on yt and really respect said that deleting his entire extensions folder with the controlnet and everything

#

and then not installing any extensions other than the newest version of the webui controlnet 1.1 and deforum and it fixed everything

#

might mess with that tomorrow

karmic crown
upbeat crane
#

cuz things are FOR sure working for some poeple, i just assume it's like an old boys network by now

#

they hold on to a lot of stuff and trade collab notebooks and stuff

#

i've never really been referencing that issue btw

#

all that was is that when 1.8 came out the deforum tabs went missing

#

thats obvi fixed

karmic crown
#

a1111 introduced some changes that many developers of extensions was now aware of

#

those extensions just disapeared

#

deforum was one of those

upbeat crane
#

for sure, it might be something interesting unfold

karmic crown
#

than some people concluded that reverting a1111 to 1.7.0 will fix the issue

upbeat crane
#

the server is wild, cuz you have all these old master level pros arguing with tons of ppl telling them they are crazy and everything is fine

karmic crown
#

but some guy found a quick solution, and deforum reapeared

#

but ONLY on 1.8.0

upbeat crane
#

and then there are tons of ppl in my shoes that saying the same thing, like 3 months ago the bottom fell out

karmic crown
#

then i notice that it just apred but not working!!!

#

so i fixed that

upbeat crane
#

and not one person can put a finger on it and say ITS THIS, DO THIS

karmic crown
#

deforum was fixed in about 24 hours

#

some people still are not aware of that, and they still sugesting reverting, that will actually now BRAKE it

upbeat crane
#

yeah the deforum dissapearing thing is recent and has nothing to do with all the stuff im referring to with it

#

you get it

#

gunna get to the bottom of it one way or another

#

but im NOT joining someones patreon to find out lol

#

like with Stable WarpFusion

karmic crown
#

i explained the whole problem in detail to one of develoeprs, and he was start fixing that imidialtely

upbeat crane
#

that guy earns all the respect tho cuz SWF is freakin amazing

karmic crown
#

we tested those changes, but i guess he need to fix few more issues

upbeat crane
#

what changes?

#

the 1.8 issue?

#

but yeah if you talk to a dev again

#

i have reinstalled 5-6 times, ive tried like probably 10-12 different control nets

#

all from different authors, different styles

#

ive thrown at least 20 different models at it

#

i can get it rendering with 1.5 models (sometimes with a single CN)

karmic crown
#

wait

upbeat crane
#

but those are always garbage compared to XL models and XL controlnets + LORAS etc

karmic crown
#

wait

#

have you using forge?

#

controlent does not work with defourm on forge

upbeat crane
#

last three days yeah almost all day long testing and troubleshooting

#

yeah, i kjnow that now lol

#

but i jumped over excitedly because controlnets broke with deforum in A1111

#

then i started seeing the stuff illysian or whatever his name is that makes them all in his github

#

yeah i was rockin n rollin for the last about 18 months before about maybe 3-4 months ago

karmic crown
upbeat crane
#

and then all of a sudden i could just no longer do anything, and have spent 90% of my time prompting and troubleshooting

#

i got some killer projects written, so once i get back up and going, ill be in heaven

#

i was looking at that like an hour ago

#

i was going to post that same page to you

#

nobody says anything outright on it, and its super obscure, kinda hard coming by

#

and to be frank, im not real sure what to even do with that

#

im gunna hit up ANDYXR here this week and see if he can walk me through some stuff

#

he is one of the mods over there, super helpful dude

karmic crown
#

anyway

#

update a1111 to latest version

#

and the update controlnet and deforum

upbeat crane
#

yeah i did

karmic crown
#

and it will work

upbeat crane
#

yeah i need to triple check on that CN1.1

karmic crown
#

controlnet should work too

#

problem with forge is that it has controlnet integrated and it also have posibility to use controlnet in some crazy places

#

and it also remove the need for some big chunks of code

upbeat crane
#

i just finished a 20 second render 30 seconds ago, with one sinngle 1.5 canny CN

karmic crown
#

but it needs more work and clean up

upbeat crane
#

but i ran about 10 or so earlier, trying xl models and cns and it would start posting those errors at the top of that page you just showed me

#

getting CUDA errors when i use XL models now, which is new

karmic crown
#

what is your GPU and what params do you have?

#

also, please check in log if you have some warnings of errors on startup

upbeat crane
#

rtx3060 12gb and ive tried almost every possible combo of parameters

karmic crown
#

ControlNet init warning: Unable to install insightface automatically. Please try run pip install insightface manually.

#

this one for example

upbeat crane
#

when i said that i just rendered a 1.5 model and a 1.5 controlnet, it actually didnt. it didnt stop rendering, it at least completed, but yeah the CN took no effect

dense tapir
#

Hello people! I would like to be informed about where a specific function is within SD, I remember using it before, however it seems as if it either disappeared, got moved, or I just forgot where it is completely.

It used to be able to extend the sides of your image and it would fill it in to not only make it larger, but give new things to it which fit the context of the image.

upbeat crane
#

when i first launch it, it looks perfect, b/c the cn's initialize etc.

#

but during each iteration, it will post "Neither the base nor the masking frames for ControlNet were found. Using the regular pipeline
2024-03-07 19:29:16,579 - ControlNet - INFO - unit_separate = False, style_align = False
*** Error running process: D:\AI\stable-diffusion-webui\extensions\sd-webui-controlnet\scripts\controlnet.py
Traceback (most recent call last):
File "D:\AI\stable-diffusion-webui\modules\scripts.py", line 784, in process
script.process(p, *script_args)
File "D:\AI\stable-diffusion-webui\extensions\sd-webui-controlnet\scripts\controlnet.py", line 1279, in process
self.controlnet_hack(p)
File "D:\AI\stable-diffusion-webui\extensions\sd-webui-controlnet\scripts\controlnet.py", line 1264, in controlnet_hack"

#

@dense tapir it's called Outpainting

#

here is a tutorial on it

dense tapir
#

Thank you!

upbeat crane
#

YW

karmic crown
#

a1111 instalations 🙈

upbeat crane
#

and so yeah ido have the latest version of SD-WebUI, the controlnet-webui extension and deforum

#

what version of python are you currently running

karmic crown
#

i have at least 4 or 5 of them

upbeat crane
#

yeah im always being told to stay with 3.10.6

karmic crown
#

but 3.10.13

#

i am on mac

upbeat crane
#

but i dont see how it could break anything now

#

so many things arent working right

karmic crown
#

so i can only use latest version of each branch

#

i have 3.9, 3.10. 3.11, 3.12, but also 3.7

upbeat crane
#

you ssaid you have 32 or 64gb of ram?

karmic crown
#

I am on macbook pro with M3 PRO

#

🙂

#

18GB of unified RAM

upbeat crane
#

i was going to have you run a test for me but since your on a mac it wont really tell me much

#

i was going to have you try like DreamshaperXL model with an XL Lora and CN with Deforum and see what happened

#

hell, id' still be interested

#

nice machine

karmic crown
#

those are all test installations

#

some of those are test of my scripts

#

and some test of deforum

upbeat crane
#

have you done any deforum rendering with XL models and CNs?

karmic crown
#

this is my new laptop, i still need to make a proper a1111 instaltion

#

not yet on this laptop

upbeat crane
#

yeah give that a shot sometime soon and hmu

karmic crown
#

just a simple tests with default model

upbeat crane
#

probably work for you, i bet

#

but yeah there are a bunch of ppl using XL models with it

#

but nobody can help the people who cant lol its weird af

karmic crown
#

my mac is little baby compare to MCutos beast Mac (M2 Ultra with 192GB of RAM)

#

some things works much faster than on many nvida cards

upbeat crane
#

is that like a pro video editing machine or something?

#

pro-tools

karmic crown
#

it is the strongest version of m2 procesor and 192GB or ram means it use up to 128GB for grapihcs

upbeat crane
#

and it cost as much as a porche

karmic crown
#

when i do a proper installation with my models, loras... we will do some test and compare how many times my m3 pro is slower than his m2 ultra

#

but, based on my initial tests, a1111 works faster than forge and comfy on this laptop

#

i have 2it/s on a1111 with 1.5 model and 1.5it/s on comfy and forge

upbeat crane
#

i get like 4.7 on 1.5 and 1.2it/s with XL models i think

#

works for me, i dont mind it at all, when im doing animation, i will just do tons of 10 second runs cuz im tweaking the camera/canvas etc

karmic crown
#

with default 1.5 model and deforum settings i get a video in about 5 minutes

upbeat crane
#

and prompt building, then at night before i go to bed, i would start a run

#

i really love the fact that you can continue broken renders with deforum

karmic crown
upbeat crane
#

such a nice feature to have. if it crashes or something happens you can easily just restart the render from the last image

karmic crown
#

3 am, time to go

#

i work in the morning

upbeat crane
#

k buddy have a good night ttyl

abstract badge
#

@upbeat crane ty for ur help the other day i have figured out control net. Turned out i didnt have the precise pair of preprocessors and processors

#

and I now have a much better understanding of it!

upbeat crane
#

that is great man, you are most welcome

#

cheers @abstract badge

dense tapir
upbeat crane
#

i honestly don't have a clue, i have never used one before

#

did you look in the video description, sebastian always puts links there

#

otherwise maybe search google for "diffusion outpainting" and check out another video or two

#

or you just just extend the commends below the video where the creator puts links and see if a different tutorial isn't linking to one

#

you can also try searching inside the extensions tab

#

that one looks pretty legit

#

comes with instructions as well

dense tapir
#

Thanks brodie I'll try

#

I don't think the inpaint model even loads in 👀 but I think I can fix this

inland stag
#

Hey! I have an issue

I was trying to generate an image, with the workflow that i got from "Area Composition Examples" that is provided by ComfyUI examples. Run the code and got this error message:

Error occurred when executing VAEDecode:

"upsample_nearest2d_out_frame" not implemented for 'BFloat16'
[And bunch of coding line error messages]

I am using SD:XL and its trained models

#

What is the issue and how can i fix it

patent carbon
hard jetty
#

Hello! Does anyone know if there is a way for me to know the mime type of my text-to-image result because I noticed that it does not have a starting string of data:<mimetype>;base64

river flame
#

Hi! Do I have to choose an embedding inside A1111 or it works anyways?

river flame
#

I m asking because it doesnt show up on my A1111

hoary plaza
#

what models and loras do yall recommend?

manic snow
river flame
hoary plaza
manic snow
#

do you want it to be realistic, semirealistic, anime or something else?

inland zephyr
#

ay I spotted these in the cmd startup, are any of them worth using?

image.png
karmic crown
#

Latest update on Deforum on A1111 issues:

Deforum now works on 1.8.0 as long as both a1111 and Deforum for webUI are up-to-date
since PR #950 is only compatible with 1.8.0, but not with legacy versions, Deforum will not work on 1.7.0 until one of the devs finishes his changes from last night and merges that. we tested those changes last night on both 1.7.0 and 1.8.0, and it worked as expected, but some further testing and tweaking is required
for those with AttributeError: 'tuple' object has no attribute 'gallery', the solution is to update a1111 to 1.8.0

GitHub
fix: modify create_output_panel output to match object returned by ...

Ever since this commit AUTOMATIC1111/stable-diffusion-webui@501993e#diff-7825c63da20059aa9e36a44157ab1a6b36a02bd71c01f5f30ea5d028320ef263R240
the signature of create_output_panel has been a class o...

hoary plaza
#

what should the order of downloading comfy ui be?

ornate elk
karmic crown
young forge
#

newbie alert: where do I find this clip skip setting on stable diffusion?

edit: nevermind. found it - Settings > User Interface > Settings in UI > add CLIP_stop_at_last_layers

image.png
hoary plaza
ornate elk
fair oxide
night orchid
#

after my understanding the automatic1111 has SD 1.8, is that correct ?

hoary plaza
ornate elk
night orchid
#

ComfyUI is not Comfy

#

I find ComfyUI confusing, but then I again I has ADHD

#

@ornate elk has you tried mov2mov ?

vocal burrow
night orchid
#

@vocal burrow you was probably replying to me

vocal burrow
#

yup mabad

night orchid
#

hehe

#

Thanks

ornate elk
night orchid
#

😦

ornate elk
night orchid
#

I cant get it to work

vocal burrow
#

let me guess, ffmpeg error ?

ornate elk
night orchid
#

no error

#

its seem to start but notting happen, notting in the cmd

#

@vocal burrow no ffmpeg error either

ornate elk
#

i say its broken, as it got updated 4 months ago

#

have you tried AnimateDIff?