🤝｜tech-support

[lost crescent]

Thanks for your help

[ruby yoke]

you're welcome

[ornate elk]

Nope not needed to upgrade RAM. But if you want to use large flux models 64gb RAM is better

Can you show the txt2img settings you used?

[spare quail]

im using img2img inpaint and i can send a full screenshot in 1 second

[mighty grove]

Can you help me generate a poster of the Black Mythical Monkey King

[ruby yoke]

wrong channel

[gilded canyon]

Is there a significant difference between a 3090 and 4090 performance wise? The VRAM is the same, but its almost a $1000 difference in my country

[ornate elk]

Yea a 4090 is much faster

But your good with a 3090 too if you just want to generate images.
Even training is fast

[gilded canyon]

I think I'll just rent a GPU on vast.ai for now

Thanks for the advice

[ornate elk]

Np

[ornate elk]

Set the preprocessor to Auto

Also check the cmd log when generating

It should show errors

[spare quail]

theres no option for auto

Error running process: /opt/rd/apps/stable-diffusion-webui-forge/extensions-builtin/sd_forge_controlnet/scripts/controlnet.py
Traceback (most recent call last):
File "/opt/rd/apps/stable-diffusion-webui-forge/modules/scripts.py", line 844, in process
script.process(p, *script_args)
File "/opt/rd/miniconda3/envs/FRGE/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "/opt/rd/apps/stable-diffusion-webui-forge/extensions-builtin/sd_forge_controlnet/scripts/controlnet.py", line 554, in process
self.process_unit_after_click_generate(p, unit, params, *args, **kwargs)
File "/opt/rd/miniconda3/envs/FRGE/lib/python3.10/site-packages/torch/utils/_contextlib.py", line 115, in decorate_context
return func(*args, **kwargs)
File "/opt/rd/apps/stable-diffusion-webui-forge/extensions-builtin/sd_forge_controlnet/scripts/controlnet.py", line 304, in process_unit_after_click_generate
input_list, resize_mode = self.get_input_data(p, unit, preprocessor, h, w)
File "/opt/rd/apps/stable-diffusion-webui-forge/extensions-builtin/sd_forge_controlnet/scripts/controlnet.py", line 164, in get_input_data
if len(unit.batch_mask_gallery) > 0:
TypeError: object of type 'NoneType' has no len()

this is what claude.ai said

[ornate elk]

Can you try with one image instead of batch?

[spare quail]

ok yeah it worked with 1 image although distored i saw the controlnet image on the right side, but whenever i do batch upload, even just 1 image, its not working at all

[vivid ridge]

Hello guys, I am working with SD upscaler and have some unclear part of it. When we use using it we use Stable Diffusion upscaler and usualy some type of GAN. How are they working like someone increase size and someone quality? Could some one explain it to me or sent some article about it? TY so much

[neon swallow]

Hello guys, we have the UI of StabilityAI/StableStudio. A few hours ago, it was working fine. However, now every new generation shows 'Error generating answer' on Stability's side. We've had this problem before, and Stability resolved it after a few hours. Do you know what's happening or why this issue is recurring? @vocal burrow @north cedar

[vocal burrow]

Everything seems to be all right on stability's api side. https://stabilityai.instatus.com/
This channel is community driven and mainly for people self hosting SD, for problems related to their online services I'd recommend contacting them directly https://stabilityplatform.freshdesk.com/support/home

[woeful knot]

Would it be a problem if I renamed a .sft file to .safetensors? I'm seeing .sft has some compatibility problems

[bleak mirage]

For what ever reason whenever I start up stable diffusion it says “Aria2 error cannot start properly closing application” I hit ok and I can still run everything despite that error. What is going on? 🤔

[ruby yoke]

it can be, yes. if you're having to do that, then you have other issues. when you updated comfy, did you update via manager or did you use the update script located in /comfyUI/update?

[ruby yoke]

can you paste your error log here please

[bleak mirage]

Sure I will see if that can be found. My error log has not been acting appropriately. I could give ya the actual error for sure 👌

[ruby yoke]

can you explain what you mean by your error log hasn't be acing appropriately?

[bleak mirage]

It would not save things pretty much. I would check but the error log would be blank 😬

[ruby yoke]

how much disk space do you have free?

[bleak mirage]

Oh easily more than a Terra bite 👌

[ruby yoke]

on your system drive

[bleak mirage]

Let me see if it’s actually working. Ooo its on another drive not my system 😮😮

[ruby yoke]

it's still going to need space on your sytem drive - and so is your sytem

[bleak mirage]

Where is the error log stored again. Its been a while 🤔

[ruby yoke]

i have no idea where it is on your sytem. but at this point i'd encourage you to look at the system error log and make sure you're not seeing any errors there.

what's your OS?

[bleak mirage]

Windows 10 👌

[ruby yoke]

https://kb.blackbaud.com/knowledgebase/Article/75433

follow those steps to get the windows event log up and go through it first

[bleak mirage]

I tried but its difficult to read. Here is the actual error👌

[ruby yoke]

okay, to start with, this is what aria2 is https://aria2.github.io/

it's a downloader.

[bleak mirage]

Yeah its tied with Civit browser I think. But that works too🤷‍♂️

[ruby yoke]

what else are you running on your sytem right now?

yeah, i'ts just being called by something in case you need to, or because that something, wants to download stuff

[bleak mirage]

Just the the stable Diffusion nothing else 👌

[ruby yoke]

which interface are you running? comfyUI? auto1111? WebUI? something else?

you might try reinstalling aria2 on your system. steps are here https://ipv6.rs/tutorial/Windows_10/WebUI-aria2/

make sure you shutdown, leave off for 10 seconds, turn on to load everything fresh before you do that and dont' run anything after you start back up till you do the reinstall

[bleak mirage]

A1111

[ruby yoke]

and what browser are you using?

[bleak mirage]

Microsoft edge

[ruby yoke]

can you switch to Chrome and see if that has issues as well?

[bleak mirage]

The error happens in the middle or ending of the bat. File start up by the way 🤔

[ruby yoke]

you're probably going to need to reinstall aria2 in that case

[bleak mirage]

Is Aria2 OS tied or is it just tied to A1111 and stable diffusion 🤔

[ruby yoke]

it's just a downloader. it's being called as you start auto1111.

you might wait for @ornate elk to log back on and see if he has any other information about the error

[bleak mirage]

Gotcha is this webUI-Aria2 a file in the A1111 same place as the bat. File? 🤔

[ruby yoke]

no idea where it's at. you can always run a search for it with windows search

[bleak mirage]

Ok 👌

[bleak mirage]

Found it Could I just try to run aria2.exe? 🤔

[ruby yoke]

sure

[bleak mirage]

It is in Civitai-browser-plus file 👌

[ruby yoke]

what is civitai browser?

and what do you use it for?

[bleak mirage]

To search for models and Loras and download them directly from the AI screen 👌

Though that works too. I could try reinstalling the extension 👌

[ruby yoke]

you could.

[bleak mirage]

Maybe corruption and such? 😬

[ruby yoke]

or a pointer got lost

[bleak mirage]

There is a file called “running” I don’t know what that could be. Despite me not running it at the moment. 🤔

[ruby yoke]

could be anything. you'd have to look through the various bat files and stuff

[bleak mirage]

Yeah 😬 but reinstalling Civit browser is a start I think 👌

[ruby yoke]

maybe. i sort of avoid chainsaw solutions if possible. IF you are going to do that, however, make sure you uninstall it all the way first, and clean up any files that it leaves behind so you really do have a fresh install

[bleak mirage]

I gotcha good point 👌

[woeful knot]

It was on forge, on the vae section it could read the clip files being .safetensors but not the vae, being .sft. It showed when I changed the extension, as far as it working, I have no idea how to spot the difference

[ruby yoke]

not sure forge can even use flux...

[pearl musk]

So I managed to improve the LoRA but it still wont work without calling for character tags, maybe it's because I'm trying to train a hairstyle lora with just one subject...

[ruby yoke]

you don't want to do that. you want a range of heads all with the hairstyle in question. and you want the images from different points of view - front, side, back, top down, etc. and since you're trying for something very specific, you need a specific tag like MyHairstyle to be used IN the prompt

and i woulid probably use around 100 images if i was training it

[pliant grotto]

how do i install this ?https://github.com/AUTOMATIC1111/stable-diffusion-webui-pixelization i dont get it

[ruby yoke]

you are probably going to want to watch a couple of how to install videos on youtube

[pliant grotto]

i didnt find anything thats specifically for this one

nvm ai came in clutch

[pearl musk]

Thanks! The problem is, the character from I want to replicate the hairystyle has just 60 images on danbooru/gelbooru/etc, all of them are full body/half body and, half of those are low quality, that is the problem I face. Maybe I should train a character lora from said char, generate images from her face/head and then use them to train a lora?

[ruby yoke]

so run them through the image upscaler under magic tools on the capcut website first

[pearl musk]

I'll try, I hope I can find images with different poses/closeups, probabably it does not help that the character has horns and a crown...

[ruby yoke]

run them all through the upscaler, then take them one at a time into photoshop and fix issues they might have, such as removing the horns and crown and so on

then if you want, you can train a lora for the character on those

generate a bunch of images, clean those up, and geneate another lora from them

[pearl musk]

Ahhh, so that was the problem too, I will have to dig up on my old photoshop tutorials

What a rabbit hole lora training is, thanks to artists drawing hairstyles with those stupid "hair intakes" or whatever they are called.

[ruby yoke]

what you're doing right now is creating a good data set. you haven't gotten to the training part. and if you don't have a good data set, your lora will not be any good, either

and once you get good images, you get to label them 😉

but the more care and the more detailed you are about everything in the set, the better the lora will be

[pearl musk]

Yeah, I agree, something I still don't understand is the "trigger word" part. Where should it go when tagging?

Is this word defined by the folder where the dataset is?

[ruby yoke]

i put the lora trigger word at the front of my prompt so i dont' forget to use it. it depends on yoru trainer, but usually, that's the keyword that is going to call up the data that's in the lora

[pearl musk]

Gotcha, thanks for the help, I will take the images I have, remove the crown and horns, and then tag them again

[ruby yoke]

welcome

[pliant grotto]

bro who tf includes an extension with no instructions piece of dogshit

after an hour managed to fix this trash, didnt even pull the launch.py correctly

nvm

new errors

godbless america

[ruby yoke]

what extension and where did you get it?

[hybrid kelp]

https://discord.com/channels/1002292111942635562/1276220594416914536

[ruby yoke]

stop spamming the channels

[rotund musk]

Is it only me or flux.1-dev fp16 isn't loading loras? I been trying to using the diffusion pipeline and pipe.load_lora_weights but I get the same results with and without lora

[ruby yoke]

are you using a lora that has a keyword?

[rotund musk]

In the LoRA page there's no mention about a keyword https://huggingface.co/XLabs-AI/flux-RealismLora , I don't know if it matters that I'm running it on Kaggle

I tried running with and without it same settings (and seed ofc) got still the same image, I know most people do it here locally but I made my own no UI Kaggle and wanted to know if it's an issue on my end, I can also send the notebook file

[flat olive]

@ornate elk when its going to be done the shark studio installation for amd?, sorry for bothering

[ruby yoke]

okay. i don't use that myself. i know that it works however, we've got it implimented for crowbot. i would ask in the Xlabs community

[mental axle]

https://huggingface.co/Linaqruf/animagine-xl-3.0

does anybody know how to install this, and if I can run it in Fooocus or if I have to get ComfyUI first?

any time I need something from github/gitlab/huggingface I can't find the download button

[ornate elk]

Download the .safetensor file from civitai.com
Then put it into models folder

[mental axle]

I don't need any of the other junk?

[ornate elk]

Nope

[mental axle]

perfect

different models interpret text differently though right

[deft hemlock]

I was able to figure out why my drive didn't work

It's because the cable was broken

I still wana try to boot sd from an external drive

I need this to happen

[rotund musk]

ah okay, i don't know if u ever used https://civitai.com/models/660136/flux64-n64-and-ps1-game-screenshot-lora but this didn't work either for me, but thx i will ask in Xlabs about that lora i sent before

if u want i can also send the notebook file in case i coded something wrong, just saying though

[true zinc]

hi guys, im super new to AI.

just trying out diff APIs now.

i tried multiple models on UIs first to see which one to use in my app that im building.

most of them were not good until i found stablediffusionweb.com.

the results are really good and i wanted ot use it.

on their website htey say they use SDXL, so i found stability.ai API offering it.

i am trying it out now with exact same prompts + params, even tried same seed - but the results are way different and ugly.

why is that? how can i just replicate exactly the same outputs as stablediffusionweb.com UI via an API? (dont mind building my own if thats what it takes but would rather use a ready API atleast for initial testing)

[hollow heath]

i want to ask again, how to unsubscribe?

when i want to update payment info it redirects me to form of new subscription

website is broken

[sonic nexus]

https://colab.research.google.com/github/Jelosus2/Lora_Easy_Training_Colab/blob/main/Lora_Easy_Training_Colab.ipynb#scrollTo=vGwaJ0eGHCkw
Does anyone know how to use this colab version?
I don't know how their connection works?
https://civitai.com/articles/4409/almost-local-lora-training-guide

[ruby yoke]

yeah, i haven't used that one either. most video game characters are fairly easy to prompt it. i'll use loras for stuff that's really unique

[ruby yoke]

unfortunately there's nothing we here on discord can do about account issues. you have to get hold of the company through their email.

[hollow heath]

what is correct email? im trying to contact since few days, with no result

[ruby yoke]

all the information i have is on this page https://stabilityplatform.freshdesk.com/support/home

just fill out a ticket and submit it

[hollow heath]

i did, 11 days ago

no reply

😠

11 days, too much for payment query

[rotund musk]

Alr

[ruby yoke]

i completely understand, but we're just the discord. we don't have anything to do with the company in the areas you need help with - we can't access your account or fix things with it. you could try pinging them on twitter perhaps

[hollow heath]

cant some community mod or dev ask someone from inside company? 😐

[ruby yoke]

you can open a ticket to the mods, but there aren't any devs around right now

you'll find the place to open tickets in the #1010934719455707218 channel

[agile musk]

Hi, im trying to Use inpaint anything, but when i install the extension through SD, It wont show up on my tabs, however it says enabled on my extension list ?

any help here?

[ruby yoke]

what interface are you using?

[agile musk]

Hi, sorry, ive managed to fix it by installing from the browser rather than directly, However now realising its incomaptible with AMD GPU's 😅

[ruby yoke]

??? stable diffusion runs fine on AMD. can you give a few more details on what is happening on your end?

[ornate elk]

You need to follow my Guides from the first link of the pinned messages.
There are all AMD Guides

[agile musk]

oh i have it all working on SD, i was talking about Inpaint anything, im currently in a process of trying to download ZLUDA

to see if this lets me use it

[ornate elk]

Yea Zluda is more compatible with extensions

Also much faster and uses less vram

[agile musk]

just a nightmare to install lol

[ornate elk]

Yea but my guide should be easy to follow

[agile musk]

brill, do you have a link to your guides please ?

nvm

i stupid

pinned

[ornate elk]

The first link 👍

[agile musk]

hmm after following the steps again, python apparently isnt installed but it is? EDIT: installing through the microsoft store aswell fixed this

okay sorry for the spam, but next issue, im stuck at this with no way to continue or loading anything EDIT: issue seemed to fix its self after 20 mins

[ruby yoke]

can you try, first, shuting down, leave off for 10 seconds to make sure anything hung in cache is cleared, then boot up to load everything fresh - and try running it again

[agile musk]

yeah ill igve that a shot as i closed it to see if it would open any faster and its stuck again!

thanks for the advice

[ruby yoke]

let's first, before any other troubleshooting make sure you're on a fresh system. let us know what happens

[agile musk]

this seemed to have fixed things!

[thorn badge]

in flux running in forge, does anyone know what i should set my "gpu weights" to? I have 16GB on my RTX 3080, and 32GB on mobo

[knotty turret]

guys how i can update my forge?

i need install everthing on github everytime?

[balmy matrix]

So, I'm having issues with my PC shutting down whenever I try and generate an image higher than 512x512 or queue multiple. The computer will shut down instantly on its own. It does not reboot. I have checked Windows Event log, it gives me nothing on why for the shut down.
I am using Windows 10, Radeon 6800XT GPU, Ryzen 5800 CPU, 32 GB RAM, 850watt PSU. I have my GPU undervolted via AMD driver software and have watched the temperatures using several programs. GPU hits temp of 100/101C before computer shut down. CPU stays around 65C.
Any idea at all as to the hardware causing this (I do presume it's a hardware issue). Is the 850Watt PSU not enough?

[ruby yoke]

that's called crashing

that sounds like memory issue

i would start by making sure you have plenty of space on your system drive and enough swap space

[balmy matrix]

I have 200 GB of free space on my system drive.

[ruby yoke]

how much paging space do you have?

[balmy matrix]

Page file space says 40 GB.

[ruby yoke]

that's probably not enough. also, make sure you're not winding up with most of the work being done on the CPU instead of all on the GPU

[balmy matrix]

That's through system info. I'm unsure how to find any more information about paging space.

[ruby yoke]

and read through @ornate elk guides for AMD in the pinned posts

[ruby yoke]

https://www.tomshardware.com/news/how-to-manage-virtual-memory-pagefile-windows-10,36929.html

[balmy matrix]

How much virtual/paged memory do you recommend?

I have read through the guides already during my own installation. I'm using ComfyUI (though before I found it I used SD.Next. Definitely prefer Comfy). The only difference from the guide is I am still using HIP SDK 5.7 instead of 6.1.2 because on the Comfy-UI github it says: "((UPDATE: HIP 6.1.2 released now, but there are problems so no need to use that one please be careful about selecting the correct version, "Windows 10 & 11 5.7.1 HIP SDK"))" So I didn't update to 6.1.2.

[balmy matrix]

I appreciate the help, btw. Spent all day on this on various githubs and reddit posts.

[runic moth]

Does anyone know how to load the old novelAI checkpoints that were just publicly released today, in comfyUI? I've loaded plenty of SD 1.5 finetunes and they always worked just fine. I'm getting an error like this:
comfyui in load_checkpoint_guess_config raise RuntimeError("ERROR: Could not detect model type of: {}".format(ckpt_path))

[balmy matrix]

Double Alexes. Alexs?

[balmy matrix]

So I increased page file space to 100 GB, virtual memory is 125 GB. PC still crashes. I have steps set to 50. It gets to about 45/46 steps, then crashes. Setting to 20/30 steps prevents PC crashing. This is the same as it was before changing to more page file space.

[ruby yoke]

yeah. you're running out of memory. how much VRAM (not ram) do you have?

[balmy matrix]

6800XT has 16 GB of VRAM

[ruby yoke]

find the pinned posts icon at the top of discord. click it. read through @ornate elk's guides for AMD

[balmy matrix]

I've already done that. :<

As stated, the only difference is I am using 5.7 AMD HIP SDK instead of 6.1, but that is because the ComfyUI-ZLUDA github said to remain on 5.7. Is that outdated now? Should I upgrade to 6.1 anyway?

[still frost]

Forge not loading, I tried to re-install by replacing all the files from the zip and running update, but it keeps getting stuck here when loading.

[balmy matrix]

I generated an image at 800x768 at 30 steps. Generated fine. I told it to generate another one. It got halfway through and crashed.

[restive sparrow]

Anyone use SDUpscale or StableSR, trying them today. keep getting IndexError: index 10 is out of range on both extensions

resetting webui fixed sd upscale but not stablsr oh well never mind that extension then'

[ruby yoke]

amd has pulled out of zluda. but i would follow the guides anyway

[ruby yoke]

you are running out of memory and it also sounds like cache isnt' clearing correctly

[balmy matrix]

What can I do to check this and fix it?

[ruby yoke]

you'll need to be running something that shows how much memory and how much GPU is being used in real time at the very least

[balmy matrix]

I have HWinfo. What exactly am I looking for, the GPU memory useage and the virtual and physical memory useage?

[restive sparrow]

nah literally juts use task manger gpu section if it starts using shared mem youre exceeding vram

[balmy matrix]

GPU temp hit 100 C right before I snipped that.

[ruby yoke]

what you want to watch for, the entire run, is how much of each is being used in real time. not just what you happen to ahve in your system

[balmy matrix]

Here's the GPU, Memory, and CPU useage at the end of a run.

GPU utilization is at 99% I noticed.

[ruby yoke]

during the run. in real time. not at the end. you want to watch how the usage is the entire time and see if you are bleeding over into shared memory at any point.

like cal said, you can open up windows taks manager, go to the GPU section, and watch how it's running while you generate

[balmy matrix]

So, during first generation, the dedicated GPU memory will hit just slightly under the maximum of 16 GB. Shared GPU memory only goes to 0.2/16 GB. When I run it again, right near the end of the step, dedicated memory shoots to max and I can see, for a single half second, that shared GPU memory goes to 2 GB/16 GB. If it goes higher than that, I don't get to see, as the computer crashes.

[ruby yoke]

so that tells you that memory isn't being cleared.

so now, i would suggest you wait for @ornate elk to log back on as he's the AMD expert, and see if he can figure out what you need to do.

work around - reboot after each run

or at least unload the program and reload it

[balmy matrix]

Thank you. Yeah, after a single run, dedicated GPU memory is stuck at 7.7 GB. Shared GPU memory is staying at 0.2, they are not decreasing.

[ruby yoke]

yup. so if you unload the program and reload it, does that fix that for the next run?

[balmy matrix]

Yep!

Gets to 7.9/16 GB dedicated GPU memory on each reoccuring "first" run. I close it, dedicated memory goes back down, open it, run it again, dedicated memory goes back to 7.7 or 7.9 GB.

I want to see if I can generate a single 1024x1024 image with dedicated gpu mem cleared.

The answer is no. It gets to about step 20, dedicated GPU gets to about 8.9/16 GB, and then it crashes. I don't see shared memory be used at all in this instance, but I presume something like that is happening.

[ruby yoke]

yeah. you probably will need to upgrade Zluda. but wait and talk to cs1o

[balmy matrix]

Will do. Appreciate all your help and patience with me.

[ornate elk]

Adjust your fan curve in AMD Adrenalin software

[ornate elk]

Your using comfyui?
What workflow?

And which model?

[ruby yoke]

🙂 you're more than welcome

[balmy matrix]

I have both comfyui and sdnext. I prefer comfyui and use the latest comfyui-zluda version. My model is pony xl. As for the workflow, I'm still slightly unsure what that means. I have a checkpoint and a vae and a lora.

[balmy matrix]

Will do.

The issue happens in sdnext as well tho. Tho you probably already knew that given it's just a gui.

[ruby yoke]

your workflow is exactly that. when you look at comfy, you see all the nodes and see how they are connected. they are doing work, and how they are connected is how the data flows from one to the other.

[fair oxide]

can 'anyone' at all explain to me why I can't inpaint or img2img with SDXL models? It always looks like this

it's becoming a hinderance on my ability to make things

it's 'only' doing this with SDXL

forget it

vae problem

damnit every time

^ me

[ornate elk]

The question is with which settings or workflow you get the crashes

For example upscaling with the wrong settings can cause that

Or generating at a to high native resolution

[ruby yoke]

yes, well - if the thing that's trying to decode isn't speaking the same langage as the thing that encoded ...

[true zinc]

hi guys, does anybody know exact model/LORAS/settings that stablediffusionweb.com uses?

for some reason their outputs are so much. better than anything i try.

[fierce hollow]

What could be the problem?

[fierce hollow]

So, I figured out that Forge needs to be updated, but I still don't understand what I need to load into these two folders to make these models work for me....

[chrome rampart]

Hope somebody can help. I transfered my local comfyai from drive D to drive F via copy. Now I get some error related to python (comfy still seems to work) F:\comfyui\ComfyUI_windows_portable>.\python_embeded\python.exe -s ComfyUI\main.py --windows-standalone-build
[START] Security scan
[DONE] Security scan
Failed to execute startup-script: F:\comfyui\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager\prestartup_script.py / Failed to initialize: Bad git executable.
The git executable must be specified in one of the following ways:
- be included in your $PATH
- be set via $GIT_PYTHON_GIT_EXECUTABLE
- explicitly set via git.refresh()
All git commands will error until this is rectified.
This initial warning can be silenced or aggravated in the future by setting the
$GIT_PYTHON_REFRESH environment variable. Use one of the following values:
- quiet|q|silence|s|none|n|0: for no warning or exception
- warn|w|warning|1: for a printed warning
- error|e|raise|r|2: for a raised exception
Example:
export GIT_PYTHON_REFRESH=quiet
Prestartup times for custom nodes:
0.0 seconds: F:\comfyui\ComfyUI_windows_portable\ComfyUI\custom_nodes\rgthree-comfy
3.4 seconds (PRESTARTUP FAILED): F:\comfyui\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager

thanks in advance EEEK seems my SD and SDForge dont run anymore when i remove old D drive (is not my bootup C)

[balmy matrix]

So remove a piece of the workflow at a time until I narrow down the cause. This is the reason my memory doesn't clear?

[ornate elk]

You can also Screenshot it so maybe we can see why it crashes

[balmy matrix]

Sure. I'll do that first. Will be home soonish and will send then. Thank you for helping.

[iron rivet]

Has anybody managed to get Flux working on a AMD 6800xt card?

[balmy matrix]

Here is my current workflow.

Okay, taking out the LORA prevented computer crashing. I was able to produce several in a row without incident. I tried switching to another LORA, that one being the Smooth Anime Lora for pony, it got to the last step and crashed. Can I not use any LORA at all? :(

16 gb of vram not enough for ponyxl plus a lora?

[gentle lava]

can someone help me

[ornate elk]

What's your GPU?

[gentle lava]

rx 580 bro

[ornate elk]

Should work with Zluda and flux 16gb model

[gentle lava]

i followed the step pinned

its using my cpu not my gpu

and i idk why

[ornate elk]

You need to follow the guide from my guides on github

First link in the pinned messages

[gentle lava]

i did it

but

show this error

[iron rivet]

the fp8 version seems to be working but its very slow... unfortunately nf4 is not working

[ornate elk]

Yep nf4 doesn't work on AMD windows cause it needs Bitsandbytes to work. That only works with AMD on Linux (or WSL) rn.

[gentle lava]

[iron rivet]

i figured, I just hope that someone manages to find a solution

i'm just a stupid end-user and have zero dev knowledge XD

[balmy matrix]

I wonder if I'll have less memory issues if I use Linux. I've not the foggiest. I do have a small Linux mint os already installed.

[gentle lava]

🤓

[iron rivet]

i once ran linux mint and it was like OS heaven!! everything was faster and nothing had memory issues

but gaming ultimately brought me back to win10

[ornate elk]

Use the sdxl fp16 fix
From here:
https://huggingface.co/madebyollin/sdxl-vae-fp16-fix/tree/main

Also you shouldn't use comfyui for zluda as the performance and vram usage is much worse than on Auto1111

[balmy matrix]

Huh. That's good to know. Is SD.Next acceptable too?

A shame as I love comfyui.

[ornate elk]

Yea but SD next does a lot different than other webuis.
Idk how good the vram usage is with it

Haven't used it in a while

[gentle lava]

[balmy matrix]

I'll take your advice and use auto1111 and use that sdxl fp16 fix.

I know using the same workflow in sd.next that I've been using in comfyui also causes my pc to crash

[gentle lava]

@ornate elk Do you at least have any idea what it could be?

[ornate elk]

Yea with auto1111 you can also use an extension called Tiled VAE which helps a lot to not crash when using Upscaling or highres

[gentle lava]

if i put no half auto1111 use my cpu

[ornate elk]

You have a path issue. It couldn't find either rocm or zluda

Both need to be in the environment path of windows

[gentle lava]

i have to put rocm and zluda on path? or i worng

i just have it

[ornate elk]

Looks good

Have you installed hip SDK rocm 5.7 ?

[gentle lava]

yes

[ornate elk]

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

Then do the extra steps from my guide needed for rx580

[balmy matrix]

For my own education: What constitutes high resolution? I can't even gen a single 1024x1024 without crashing. Or two 864x768.

[gentle lava]

I put hip path and zluda only in the below path, do I have to put it in both?

[ornate elk]

Thats very strange.
But can you try to setup auto1111 with Zluda first. Its easier for me to compare and troubleshoot

Then we can test

[balmy matrix]

Yessir. Will do. Whatever helps you to help me.

[ornate elk]

No only in the bottom one

[azure magnet]

Hi, could someone help me with Ipadapter. Should i put CLIP-ViT-H-14-laion2B-s32B-b79K
CLIP-ViT-bigG-14-laion2B-39B-b160k both off this two models in the clip_vision folder?
And how should i rename those?

[gentle lava]

right?

[gentle lava]

Additional Step ONLY for RX580 Users:
Go into the stable-diffusion-webui-amdgpu folder and click in the explorer bar (not searchbar)
There Type cmd and hit enter.Then type and run these three commands on by one:
venv\scripts\activate
pip uninstall torch torchvision torchaudio -y
pip install torch==2.2.1 torchvision==0.17.1 torchaudio==2.2.1 --index-url https://download.pytorch.org/whl/cu118

this?

[ornate elk]

Exactly

[gentle lava]

okay

[balmy matrix]

Got auto1111 installed, but I'm going ahead and updating to Rocm6.1 since I was still on 5.7, and the install guide says to do that anyway, I hadn't before because the comfyui-zluda github said to stay on 5.7.

[chrome rampart]

Hope somebody can help. I transfered my local comfyai from drive D to drive F via copy. Now I get some error related to python (comfy still seems to work) F:\comfyui\ComfyUI_windows_portable>.\python_embeded\python.exe -s ComfyUI\main.py --windows-standalone-build
[START] Security scan
[DONE] Security scan
Failed to execute startup-script: F:\comfyui\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager\prestartup_script.py / Failed to initialize: Bad git executable.
The git executable must be specified in one of the following ways:
be included in your $PATH
be set via $GIT_PYTHON_GIT_EXECUTABLE
explicitly set via git.refresh()
All git commands will error until this is rectified.
This initial warning can be silenced or aggravated in the future by setting the
$GIT_PYTHON_REFRESH environment variable. Use one of the following values:
quiet|q|silence|s|none|n|0: for no warning or exception
warn|w|warning|1: for a printed warning
error|e|raise|r|2: for a raised exception
Example:
export GIT_PYTHON_REFRESH=quiet
Prestartup times for custom nodes:
0.0 seconds: F:\comfyui\ComfyUI_windows_portable\ComfyUI\custom_nodes\rgthree-comfy
3.4 seconds (PRESTARTUP FAILED): F:\comfyui\ComfyUI_windows_portable\ComfyUI\custom_nodes\ComfyUI-Manager

Comfy UI does seem to work tho

[gentle lava]

venv "C:\AMD\SD-Zluda\stable-diffusion-webui-amdgpu\venv\Scripts\Python.exe"
WARNING: ZLUDA works best with SD.Next. Please consider migrating to SD.Next.
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-2-g395ce8dc
Commit hash: 395ce8dc2cb01282d48074a89a5e6cb3da4b59ab
Using ZLUDA in C:\AMD\SD-Zluda\stable-diffusion-webui-amdgpu.zluda
Skipping onnxruntime installation.
You are up to date with the most recent release.
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
C:\AMD\SD-Zluda\stable-diffusion-webui-amdgpu\venv\lib\site-packages\pytorch_lightning\utilities\distributed.py:258: LightningDeprecationWarning: pytorch_lightning.utilities.distributed.rank_zero_only has been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it from pytorch_lightning.utilities instead.
rank_zero_deprecation(
Launching Web UI with arguments: --use-zluda --update-check --skip-ort --medvram

rocBLAS error: Cannot read C:\Program Files\AMD\ROCm\5.7\bin/rocblas/library/TensileLibrary.dat: No such file or directory for GPU arch : gfx803

rocBLAS error: Could not initialize Tensile host:
regex_error(error_backref): The expression contained an invalid back reference.
Pressione qualquer tecla para continuar. . .

show this error now

[ornate elk]

Make sure you replace the zluda files and also check the environment paths

[ornate elk]

Check out this:
https://github.com/CS1o/Stable-Diffusion-Info/issues/2

[balmy matrix]

Getting this error when I launch webui-user. I'd have posted a log, but I couldn't find one anywhere, so apologies if it's too large or annoying.

Sorry, meant to post this one first.

[ornate elk]

You need to have Git installed

And also python 3.10.11

[balmy matrix]

I do have git installed. That's weird. I'll install the older version of python.

[ornate elk]

After that delete the venv folder and relaunch the webui-user.bat

[iron tartan]

😉

[gentle lava]

what i have to do with these

@ornate elk

[balmy matrix]

Made a mistake earlier saying I removed the vae, when instead I had removed the lora. Have edited that.

I've got automatic1111 up and working now, about to test with just the vae you gave me and nothing else. I have also enabled "hypertile vae"

Should I also enable hyptertile u-net?

[balmy matrix]

Some good news: with the vae you have and hypertile vae enabled, I was able to generate a 1024x1024 image at 20 steps. I tried again at 30, then at 40, succeeding. I tried to 50 steps, it got to about 80 percent through, then the pc crashed.

[ornate elk]

Replace the rocm library files with either one of these and place the DLL like I mentioned

[ornate elk]

No don't use hypertile. I meant the tiled diffusion extension which has a tiled VAE setting.

[balmy matrix]

ah, gotcha

[ornate elk]

And yes use the sdxl fp16 VAE for sdxl that helps a lot

[balmy matrix]

I did.

[ornate elk]

Also make sure to close every other GPU heavy programms like games or Wallpaper Engine

[balmy matrix]

Still, you'd think I could produce a 50 step 1024x1024. I know 50 isn't necessary, but I thought it'd be a simple measure to test.

I have nothing open except my browser, discord, and automatic1111

[ornate elk]

Yep should work without problems at a normal 1024x1024

[balmy matrix]

Except it crashed at 50 step 1024x1024

I click the "Tiled Diffusion" check box now, right?

[ornate elk]

Only check tiled vae

And leave it at default values

[balmy matrix]

Got it.

Time to add a lora and see what happens

[ornate elk]

Also you can try add --medvram-sdxl to the webui-user.bat and test if that helps

[gentle lava]

i tried

bro

[vast pulsar]

Hey guys, in webui I was told --medvram would help speeding up the process but... Where exactly on the bat do I type this and what do I type exactly?

Cuz like the console keeps telling me it's launching with --lowvram which is not the case for my machine

Got 16 of hem gb to spare

[gentle lava]

venv "C:\AMD\SD-Zluda\stable-diffusion-webui-amdgpu\venv\Scripts\Python.exe"
WARNING: ZLUDA works best with SD.Next. Please consider migrating to SD.Next.
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-2-g395ce8dc
Commit hash: 395ce8dc2cb01282d48074a89a5e6cb3da4b59ab
Using ZLUDA in C:\AMD\SD-Zluda\stable-diffusion-webui-amdgpu.zluda
Skipping onnxruntime installation.
You are up to date with the most recent release.
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
C:\AMD\SD-Zluda\stable-diffusion-webui-amdgpu\venv\lib\site-packages\pytorch_lightning\utilities\distributed.py:258: LightningDeprecationWarning: pytorch_lightning.utilities.distributed.rank_zero_only has been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it from pytorch_lightning.utilities instead.
rank_zero_deprecation(
Launching Web UI with arguments: --use-zluda --update-check --skip-ort --medvram
Pressione qualquer tecla para continuar. . .

🤡🤡🤡🤡🤡🤡🤡🤡

i did this like 8x

[vast pulsar]

Skill issue

[gentle lava]

what?

[ornate elk]

Restart the PC and relaunch

[gentle lava]

i did it 3x bro

i will do again

[vast pulsar]

Do it 4 more times and you unlock a secret level

[chrome rampart]

hey all. Since i seem to have Python issue is it smart to de-install and do new install or risks ?

[gentle lava]

will I get a photo of your round butt???

[vast pulsar]

If you tell me where and how I type that medvram thing then maybe 🥵

I am the real John fallout afterall

War doesn't change

[gentle lava]

webui user

lol

[vast pulsar]

No, I mean where and how exactly

I know. It's in that bat

[gentle lava]

webui-user> right click> edit> type "--medvram" on (COMMANDLINE_ARGS=)

[vast pulsar]

Anywhere?

Not under some other bs?

aaaah

Okie

Thank you, you saved the post apocalyptic wasteland

[gentle lava]

put a space at the beginning

after that click in file and save

or ctrl + s

[balmy matrix]

@ornate elk You're a damn brilliant man, fella, thank you. Following everything you've given, I am using the VAE you gave, a lora along with it, tiled VAE, and I'm able to produce 1024x1024 images without crashes. I can even produce them via batch count (though I won't dare to produce via batch size). Finally. It is greatly appreciated, truly. :)

My shared gpu memory doesn't even get used. 👍

[gentle lava]

@ornate elk idk what to do anymore

[vast pulsar]

time to test this bitch of a bat out

should i delete

lowvram?

@gentle lava

[gentle lava]

yeah

[vast pulsar]

aight

[ornate elk]

Perfect 🙂 glad it worked.
Comfyui has a similar node for tiled VAE. Maybe you can experiment with that.

[ornate elk]

Is that the whole log? Does the webui opens?

[gentle lava]

yeah

it open

i go to website too

but its dont generate

and after a while the site dont work anymore

wait

[ornate elk]

What does the CMD shows after pressing Generate?

[gentle lava]

@ornate elk

nothing

I thought I had to wait some time for it to start generating but I waited about 15 minutes and nothing happened

and this appear after a while

[ornate elk]

What does the task manager shows when generating?

[boreal python]

ever since i used pony diff xl as checkpoint, my image generation time has increased drastically. any idea how to eliminate this lag? what does that A, R, Sys stand for in the right side of the image

[gentle lava]

i will see

[karmic talon]

I'm using Automatic1111. I did a git update and now I get 'Found no NVIDIA driver on your system'. But I can run LLMs just fine? It seems just StableDiffusion is broken for me

[ornate elk]

Pony is sdxl based means that its also trained on a native resolution of 1024x1024

512 won't work

Also you should use xformers if your on nvidia

[ornate elk]

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

[gentle lava]

almost 4gb

[boreal python]

so should i download sdxl ? how do i use xformers ?( i do have NVidia)

[karmic talon]

ok trying. I use webui.sh btw. webui-user.sh just returns to command prompt with no error inmediately

now I get 'launch.py: error: unrecognized arguments: COMMANDLINE_ARGS = ' after webui.sh does its downloads

[paper crag]

say, i noticed some of the loras i got in my folder arent appearing when i search for them in SD.
This happened to any of you?

[karmic talon]

ok I fixed the above error. Now I get 'RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check'.

[ornate elk]

Are you on Linux ?

[ornate elk]

Edit the webui-user.bat
At the line commandline_args=
You add: --xformers

Then save and relaunch

[ornate elk]

Make sure they are for the right model version

Or they won't show up

[boreal python]

yeah i just did thanks

[gentle lava]

when i generate a image its start to generate now but still in %0 and show this error

ITS WORKING

OMFG

[chrome rampart]

Anybody tip/advice >Since i seem to have Python issue is it smart to de-install and do new install or risks ?

[karmic talon]

yes on ubuntu. Worked before. And no driver change

[ornate elk]

Nice ^^

[gentle lava]

OutOfMemoryError: CUDA out of memory. Tried to allocate 3.52 GiB. GPU 0 has a total capacity of 8.00 GiB of which 3.46 GiB is free. Of the allocated memory 2.15 GiB is allocated by PyTorch, and 744.96 MiB is reserved by PyTorch but unallocated. If reserved but unallocated memory is large try setting PYTORCH_CUDA_ALLOC_CONF=expandable_segments:True to avoid fragmentation. See documentation for Memory Management (https://pytorch.org/docs/stable/notes/cuda.html#environment-variables)

i have to use lowvram? @ornate elk

[ornate elk]

If --medvram doesn't help, yes

[gentle lava]

if i use lowvram the image get worse or this just help the auto1111 no use so much ram?

sorry my english is poor

XD

[karmic crown]

image quality does not depend on a GPU, memory, or any other hardware part
it depends ONLY on MODEL and settings you are using

[gentle lava]

i know but if i use --lowvram the image get worse?

[karmic crown]

read wha i wrote

settings = IMAGE SETTINGS

not memory settings

[gentle lava]

i think i understad

[karmic crown]

you can create better image with CPU only than with 4090
if you use better model, and better settings

with same settigns and model only difference is TIME

low-end GPU with low memory = SLOW
fast GPU with a lot of memory = FAST

choose a decent model from civitai, do not use default models
use the recommended number of steps, scheduler, and dimensions...
and you will get a good image, no matter how slow your graphics is

[gentle lava]

i tried to generated and..

NotImplementedError: Cannot copy out of meta tensor; no data!

[gentle lava]

oohhH

thanks bro ❤️

[karmic crown]

i had a chance to try automatic on nvidia a100 with 80GB
hundreds of images in seconds

but quality is the same as on 1660

[gentle lava]

@ornate elk

[karmic crown]

what is your gpu?

[gentle lava]

rx 580

i just put a new model and show this error

[karmic crown]

show me your webui-user.bat

[gentle lava]

@echo off

set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS= --use-zluda --update-check --skip-ort --lowvram
call webui.bat

[karmic crown]

looks fine

if you get ANY error related to memory, always close cmd first, and then re-run

which model you downloaded? what setting did you use?

[gentle lava]

i changed lowvram to medvram and its working

[gentle lava]

counterfeit

[gentle lava]

(masterpiece, best quality),1girl, solo, flower, long hair, outdoors, letterboxed, school uniform, day, sky, looking up, short sleeves, parted lips, shirt, cloud, black hair, sunlight, white shirt, serafuku, upper body, from side, pink flower, blurry, brown hair, blue sky, depth of field
Negative prompt: EasyNegativeV2
Steps: 25, CFG scale: 10, Sampler: DPM++ 2M Karras, Seed: 2311942344, ENSD: 31337, Size: 1024x512, Model: CF5_Counterfeit-V3.0_fix_fix_fix, Model hash: db6cd0a62d, Hires upscale: 2, Hires upscaler: R-ESRGAN 4x+ Anime6B, Denoising strength: 0.45, Clip skip: 2

[ornate elk]

Make sure to set hires steps to 10 or it can crash

Or freeze

[karmic crown]

CF5_Counterfeit-V3.0_fix_fix_fix ?!

[gentle lava]

yes

[paper crag]

thats the confusing part, i'm using pony checkpoint yet they not appearing in spite of being pony loras.

[gentle lava]

fp32

[karmic crown]

i am not sure if you can run fp32 model with --lowvram

[gentle lava]

i put medvram and now work lol

run out memory

i will try fp16

[karmic crown]

smaller model = lower need for ram

easy 🙂

[gentle lava]

majicMIX realistic 麦橘写实 is 2gb and auto111 said its out of ram

[karmic crown]

and you used 512x1024 with 4x upscale?

that is not a small amount of ram 🙂

i am not sure if tiledvae can work and help with your config, CS1o knows that

[gentle lava]

i have to use 512x512 then?

or 600

idk

[karmic crown]

try without upsacle first

than if it work, try x2

than see what happens

Hires upscale: 2

you can try with 1.5

[gentle lava]

(masterpiece, best quality),1girl, solo, flower, long hair, outdoors, letterboxed, school uniform, day, sky, looking up, short sleeves, parted lips, shirt, cloud, black hair, sunlight, white shirt, serafuku, upper body, from side, pink flower, blurry, brown hair, blue sky, depth of field
Negative prompt: EasyNegativeV2
Steps: 25, Sampler: DPM++ 2M, Schedule type: Karras, CFG scale: 10, Seed: 402044191, Size: 512x1024, Model hash: cbfba64e66, Model: counterfeitV30_v30, Version: v1.10.1-amd-2-g395ce8dc

Time taken: 1 min. 57.2 sec.

[karmic crown]

why your cfg is so big?

CFG scale: 10

[gentle lava]

its from the original image from the model

what i put? 5?

[karmic crown]

i usually start with 7

check images below

some has 3.5, 5, 7....

10 is too much

[ornate elk]

Always set hires steps to 10

When upscaling

[karmic crown]

0 = same as generate?

i forgot

[gentle lava]

okay i put

i will generate

[ornate elk]

If it won't work you need to change the Upscale by to 1.3 or install the tiled vae extension
Also dont use native 1024 resolution with 1.5 based models.
Try 512x768 for example.

I'm off now

[gentle lava]

out of ram again

[gentle lava]

okay

i will try again

[flat olive]

it just crash should i reinstall?

[ornate elk]

Have you followed my guide?

[flat olive]

yep

[placid pier]

Updated my nvidia the other day, today I get this error: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check
I haven't done anything besides updating and closing my PC, to come back to this today.

[ornate elk]

What does the cmd and taskmanager shows when you try to generate?

[agile musk]

So, the other day, i was generating images pretty fast, im trying today, and its getting stuck at 94% for about 5 minutes, per image, any suggestions, ive changed no settings at all since then

[solemn moon]

hey bro's, will flux go to Auto, even eventually..?

[karmic talon]

Auto1111 recognizes two loras out of of a dozen I downloaded. How come

[karmic talon]

ok found it it's in the settings

[solemn moon]

gosh im panicking

[ruby yoke]

you'd probably get better results asking this on the flux github...

[solemn moon]

bump a link bruvs

openness is what we're abt right? it what we're abt right?

[ruby yoke]

here's the auto1111 main github page https://github.com/AUTOMATIC1111

[ruby yoke]

yeah, but you're asking a design question that only the a1111 devs can answer. all we can do here is guess

[solemn moon]

sr i read that as "discord" my bad

[umbral compass]

Trying out Forge UI, followed the installation guide. I ran the "run.bat" file, but do I also need to run "webui-user" file? It seems to be the same thing and only thing is it looks like it installed couple of files or something. (I'm using the f0017 ver.)

[umbral compass]

I'm also getting error for Forge, I'm running 3.10.11 Python (Like I do with ComfyUI and A1111) with no issues. But with Forge I can't get it to load.

If however, I do "@REM" it would work fine, but I won't be able to have my other models linked with A1111. How am I suppose to get it to work?

[flat olive]

all that goes before the crash, and the taskmanager just shows the use of gpu

if it's that what you mean

[placid torrent]

@ornate elk Can I dm you

[drifting dirge]

hello id like to ask for some gpu advice.

do you guys think upgrading to an ASUS Geforce RTX 3060 Dual OC V2 for 322$ is a good idea or what would you recommend ||(also saw a MSI GeForce RTX 3060 VENTUS 2X OC 12 GB
for 362$)||? primarily id like to do image gen (maybe some light lora training as well). currently my specs are:
`* Gigabyte H77-D3H

• I7 3770k 3,5 ghz
• two DDR3 8gb 1600hmz samsung, PC3-12800(16gb)
• Geforce GTX 760 HAWK 2 gb DDR5 (1111Mhz)
• Corsair CX600 80-plus bronze 600W`

[ruby yoke]

i recommend an intel cpu and the best nvidia gpu you can get

[drifting dirge]

i do have an intel CPU but idk if those GPUs are good for this?

[drifting dirge]

ive heard the same thing but ive heard some nvidia gpus (even though they have sufficient vram) dont work properly with SD 🤔

do you think any of those listed GPUs would work? i saw some ppl on the SD reddit praise the 3060s 2 years ago but idk if theyve become outdated? ideally id also like to be able to train on my own digital paintings

gotcha gotcha. thanks for your input. i suppose ill have a look at some price history and such to figure out the pricing part.

[late sapphire]

can yall help me install SD?

nvm guys i figured it out

[thorn pendant]

can anyone guide me those thing

[ruby yoke]

you just select what of those you want

[stark parrot]

I WAS USING FORGE FINE . AFTER I UPDATED I GOT THAT ERROR

I USED THE UPDATE CMD WINDOW

[solemn moon]

does any1 know of a Control for Flux Auto?

[sonic ice]

i was doing jsut fine making images, but now no matter what lora i use or what i do the faces are messed up. but i didnt change anything

[fierce hollow]

Who has worked with ControlNet? Please help me, I have tried to adjust all the settings (I mean the sliders ‘Control Weight’ and ‘Timestep Range’). I generated a girl of 25 years old, and I want her face again, but for some reason I can not, please tell me what I do wrong ?
https://imgur.com/a/bVvQBwJ

[flat olive]

Is this output image somenthing?

[ornate elk]

When do you get that ?

[ornate elk]

It has to look like this:
#🤝｜tech-support message

[ornate elk]

Can you show an example + txt2img settings?

[flat olive]

using a1111+zluda, just generating

[ornate elk]

Which model?

Can you try an other model and without embeddings?

[flat olive]

@ornate elk i was opening it and it just crashed

is it normal?

[ornate elk]

Your webui is not updated

[flat olive]

yeah i updated just a few seconds i will try again

it works now but it wasn't until i updated, the update box warning, just don't showed the last time

👍

[ornate elk]

Ah okay, good

[serene lake]

How long does it typically take for an payment to go through? I payed 10min ago but I'm still getting 402 responses from the API working now so I guess around 15-20min

[raw vine]

Does any guys know why my Depth and segmentation control pops out error when I try to see the preprocessor preview?

[ornate elk]

Can you show the Controlnet depth settings ?

[ornate elk]

So just the preview doesn't work?

[raw vine]

No, the image that I generate doesn’t get the control from it too

The one you see rn is the image from me using canny

[fierce hollow]

help..

[desert grove]

Howcome sometimes just 1 word in the prompt can cause this?

[vocal burrow]

settings ?

[desert grove]

What in settings?

Usually it'll occur randomly and it'll be because of my current prompt configuration, and I'll have to remove or shift some words around to get it to not do that.

[vocal burrow]

what settings are you using to produce such images, models, prompts, sampler, etc... ? basically screenshot your whole webpage

[desert grove]

Is this what you're asking for?

[vocal burrow]

yes. I see nothing really sus here. It could be the sampler's fault.

If you can reproduce it, try with a different sampler.

Also what command line arg do you have ?

[desert grove]

Alright

Also, command line arg?

[vocal burrow]

what s in your webui-user.bat ?

(drop it in here)

[desert grove]

@echo off

set PYTHON=C:\Users\Archi\AppData\Local\Programs\Python\Python310\python.exe
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--theme dark

call webui.bat

[vocal burrow]

and what s your gpu ?

[desert grove]

RTX 3060 ti

[vocal burrow]

ok, it s not gonna fix your issue. but I suggest adding --xformers to the set set COMMANDLINE_ARGS=--theme dark line. it will make things faster.

let me know if you find steps to reproduce the issue and I ll see if I can reproduce them too on my end (later tonight)

[desert grove]

Alright, thank you.

[trail pumice]

guys my upscale model doesnt work correctly

i assume this is the problem

can anbody tell me how to solve this?

[ornate elk]

You need --xformers --medvram-sdxl

Also you need to update your webui

[ornate elk]

Would need to see the full cmd log

[trail pumice]

i am not using webui btw

it is comfyui

[ornate elk]

Oh okay

[ornate elk]

Whats the upscaler file ending?
.PTH or .safetensors?

[trail pumice]

pth

[ornate elk]

Have you tried other upscalers?

[trail pumice]

yes i did

they are giving same warnings

this upscalers just increasing image size

doesnt work in usual way to be suppose

like this model have restoration denoise and sharpening

but i cant see this results

[hot belfry]

tried to use mimic motion but having black generation. i am losing my mind help would be appreciated. thanks.

[hot belfry]

please : )

[sonic ice]

yes

[solemn moon]

im gonna give this a try, assuming your add detailer 1 is the same as 2
you should add what you want to see, like "beautifull face" (and negatives)

enabled but empty

i could very well be wrong

[sonic ice]

yes. enabled but empty. 1 is yolo hand the other is ylo face

[solemn moon]

im busy using ad detailer

beautiful face, intricately decorated, soft smooth lighting, soft colors, 100mm lens, 3d blender render, trending on polycount, modular constructivism, physically based rendering, centered, volumetric light, ultra realistic

blurry, toy, underwater, photoshop, out of focus, scary, creepy, evil, disfigured, missing limbs, ugly, gross, missing fingers, text, cropped, barcode, eyewear, fantasy

[solemn moon]

give it a try, if not, start deleting loras

[knotty turret]

can someone help me idk what to do

[solemn moon]

torch is not compiled with cuda

[knotty turret]

how to solve?

[sonic ice]

already tried that

[knotty turret]

tried to reinstall comfyui alot times but this keep give me this error

[knotty turret]

already tried it

[solemn moon]

maybe set vae to automatic

that fuzzy img looked like vae problems for me b4

[solemn moon]

gpt?

[karmic crown]

show me your webui-user.bat

what is your GPU?

can you run webui at all? if you do, please screenshot and post the bottom where it listed versions of webui, torch...

[knotty turret]

Its comfyui

[karmic crown]

ok

[karmic crown]

its not important for comfy

i thougt its a1111

anyway

show me the whole output you are gettings

[knotty turret]

ok

[karmic crown]

copy paste whole text from cmd

[knotty turret]

can you read? or you need text?

i can copy it too

[karmic crown]

its fine

you have python 3.11 x64, so that is fine

did you use venv or directly installed requirements?

[knotty turret]

just download it from github using the portable nvidea and after that i installed the Animate Anyone because i want to use that. after i install the missing nodes that error happen

trying my best to explain in english 😅

[karmic crown]

its fine

english is my 3rd language 🙂

[solemn moon]

yr english is good 😄

[knotty turret]

thx

[karmic crown]

portable should have python somewhere inside it

let me check

[knotty turret]

ok

[karmic crown]

i am on mac, so i need to run virtualmachine, and then find and download it 🙂

[knotty turret]

ok

i mean if u want screenshot too

can send too

[karmic crown]

there is a new version

[knotty turret]

i using that one

[karmic crown]

ok

its slow for my connection, it should be double

but its fine

[knotty turret]

yeah np can wait

my upload for some reason cant pass 36mbps idk how to fix that

tried alot things

[karmic crown]

its stoped like this

please wait

27MB/s is around 240Mbps, and i have 600 Mbps

[knotty turret]

i have the same problem

have 500 Mbps

[karmic crown]

ok

[knotty turret]

😃

[karmic crown]

so, you should have a folder python_embeded inside this

[knotty turret]

yeah i have it

[karmic crown]

give me a second, it need some time to extract, because my virtual disk image is dyniamc, so it need more time

[knotty turret]

I need to go to the bakery but I'll be back soon

[karmic crown]

it looks like this version does not have pip, and ensure pip does not work

but this do work

python_embeded\python -m pip --version

so we can use that to try to fix your installation

python_embeded\python -m pip install -r ComfyUI\requirements.txt

you have to run it from that folder that contains those 3
it should be ComfyUI_windows_portable (if you havent change the name)

if that doesnt help, try

python_embeded\python -m pip install torch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1 --index-url https://download.pytorch.org/whl/cu121

also, check if you have cuda 12.1 or 11.8

if you have 11.8, you need another torch

python_embeded\python -m pip install torch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1 --index-url https://download.pytorch.org/whl/cu118

ensure that you are in right folder, for example

or wherere you installed it

i just tried all commands i sent you

and they work as intended

[knotty turret]

i back

alright let me try

@karmic crown

[round ravine]

running SD on AMD Gpu for first time and waiting in this screen for 10 minutes

Do I need to wait more?

[knotty turret]

it should be 12.1 cuda?

[karmic crown]

evne the lagtest version of torch does not support that

you need 12.1

[knotty turret]

how to downgrade?

[karmic crown]

i am mac user, dont ask me

🙂

[knotty turret]

oh ok

@karmic crown wait

tried this command python_embeded\python -m pip install torch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1 --index-url https://download.pytorch.org/whl/cu121

and it worked

lol

but there just one thing that gives a error now

something about xformes

a error python pop on my screen but i cant take screentshot

and then comfy open

[karmic crown]

xformers are aslo nvidia thing, we do not have xformers, nor cude here 🙂

[knotty turret]

can just ignore it?

[karmic crown]

i am not sure

depends what it says

[knotty turret]

want see?

[karmic crown]

sure

[knotty turret]

let me get my phone

[karmic crown]

but i would try to change cuda to be 12.1

someone else will help later i guess

i can help with pyton, but not su much with windows stuff

[karmic crown]

huh

Unable to locate the entry point of the
procedure

i would not ignore that

https://github.com/bycloudai/SwapCudaVersionWindows

check this, you should be able to change it

[knotty turret]

ok

11.8 its the correct one right?

why on my nvidea-smi show 12.6

[karmic crown]

wait

[knotty turret]

ok

[karmic crown]

wait

it should be cu118 i guess

let me check

[knotty turret]

alright

[karmic crown]

CUDA 11.8

python_embeded\python -m pip install torch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1 --index-url https://download.pytorch.org/whl/cu118

i am sorry for a typo

[knotty turret]

use this?

[karmic crown]

yes

[knotty turret]

ok

[karmic crown]

this one is fine

2.3.1 with cuda 118

[knotty turret]

already satisfied

let me open again

same error pop on screen

[karmic crown]

ok

than do this

python_embeded\python -m pip uninstall  torch torchvision torchaudio

than install with that line

[knotty turret]

ok

[karmic crown]

wait, your xformers is a proble, not this, but ok

that will not hurt

[knotty turret]

do anyway?

[karmic crown]

python_embeded\python -m pip uninstall xformers

[knotty turret]

both?

[karmic crown]

it would be best

[knotty turret]

ok

ok what now

python_embeded\python -m pip install torch torchvision torchaudio

this right?

[karmic crown]

wait

[knotty turret]

ok

[karmic crown]

python_embeded\python -m pip install torch==2.3.1 torchvision==0.18.1 torchaudio==2.3.1 --index-url https://download.pytorch.org/whl/cu118

this

[knotty turret]

ok

[karmic crown]

than try to run comfy

[knotty turret]

alright just finishing the download first

its stuck 🥺

let me run again that command

[karmic crown]

it might need some time, even your connection to internet is fast, the connection to github might not be at this moment

[knotty turret]

its working better now

[trail summit]

Hi did anybody have this error with the API?

Error: API error: 400 {"errors":["image: is required"]

I'm using it with JS exactly like in the docs, I have no idea why it fails. Someone here posted that adding filename to the image helps but I couldn't make it work 😦

[knotty turret]

will do it now

no error

this time

[karmic crown]

most people here use local installed apps, like stablediffusion webui, comfyui...
but i hope someone will know the answer, maybe it is better to write to official support, since this is unofficial (comunity) channel

[knotty turret]

just opened normal

gg?

[karmic crown]

if there is no error, it should be fine

[karmic crown]

try to generate something

[knotty turret]

i dont have any workflow

🙃

nvm i have

[karmic crown]

it looks fine to me

[knotty turret]

lol i need model first

i dont have any

[karmic crown]

what is your gpu again?

[knotty turret]

2060

[karmic crown]

you can try this model (i like it a lot)
https://civitai.com/models/200061/hephaistosnextgenxl-lcmbf16hybrid

[knotty turret]

alright

vae?

[karmic crown]

try with automatic

[knotty turret]

ok

its a realistic model?

can genarate anything? like Pony V6

[karmic crown]

"!USE DPM++ SDE/Karras or DPM++ 2Sa/Karras!"

this is from the author

[knotty turret]

ok

dpm++ its dpm_2 ?

[karmic crown]

dpmpp_sde

[knotty turret]

ok i found it on reddit

[knotty turret]

yeah

[karmic crown]

dpmpp_2s_ancestral

those two

[knotty turret]

classic one image

thx u man

everything see to work fine for now😅

btw do u have any good workflow?

[buoyant palm]

Hi. My controlnet does not work properly. It is the last version. Any forum where i can check what is going on? thanks

[karmic crown]

i am using a1111 mostly

i do not remember when i used comfy last time

but to be honest, i lately spent much more time generating music than images 🙂

[knotty turret]

ok

[karmic crown]

you need to be sure that you are using the right combination of base and controlnet model

XL base + XL controlent model
or
1.5 base model + 1.5 controlnet model

you cant mix

everything that is not 1.5, basically should use xl controlnet, but i am not sure if it works for all (pony for example)

i have never used pony model, so i have no idea

[buoyant palm]

ok, thank you. I use XL. Maybe that is the problem.

chaz gpt tellsme it could be gradio. I will also check that.

[karmic crown]

chatgpt is not good for that

[gentle lava]

Do you have any idea what eta means? I see some prompts on civitai and there is something related to "Eta: 0.2" or something like that

[karmic crown]

its great for some things, but have no clue about ai image generation

[buoyant palm]

where can i find the controlnet xl models?

[karmic crown]

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

there is "one-in-all" also

i am trying to find it

https://huggingface.co/xinsir/controlnet-union-sdxl-1.0

[buoyant palm]

thank you so much 🙂

[gentle lava]

How do I add %HIP_PATH%bin to the path?

[oblique carbon]

In comfy, there are a few different nodes related to tiled ksampling, supposedly allowing you to work on very large, detailed images. I haven't had much luck with them so far. Can anyone direct me to a workflow for this? Or explain the difference between these nodes? There's one from BlenderNeko, another from FlyingFireCo and a Tiled Diffusion & VAE from shiimizu. I tried to ask co-pilot what the difference was: you can imagine how useful that was. 🙄

[gentle lava]

i have to press "new" type "%HIP_PATH%bin" ?

[ruby yoke]

what OS are you running

[gentle lava]

windows

@ruby yoke Could I put it in, help me with something?

venv "C:\SD-Zluda\stable-diffusion-webui-amdgpu\venv\Scripts\Python.exe"
WARNING: ZLUDA works best with SD.Next. Please consider migrating to SD.Next.
Python 3.10.11 (tags/v3.10.11:7d4cc5a, Apr 5 2023, 00:38:17) [MSC v.1929 64 bit (AMD64)]
Version: v1.10.1-amd-4-gb0d9eb6d
Commit hash: b0d9eb6df1f6631a49988a9f705ff568f908aa2b
Failed to load ZLUDA: Could not find module 'C:\SD-Zluda\stable-diffusion-webui-amdgpu.zluda\nvcuda.dll' (or one of its dependencies). Try using the full path with constructor syntax.
Using CPU-only torch
Skipping onnxruntime installation.
You are up to date with the most recent release.
no module 'xformers'. Processing without...
no module 'xformers'. Processing without...
No module 'xformers'. Proceeding without it.
C:\SD-Zluda\stable-diffusion-webui-amdgpu\venv\lib\site-packages\pytorch_lightning\utilities\distributed.py:258: LightningDeprecationWarning: pytorch_lightning.utilities.distributed.rank_zero_only has been deprecated in v1.8.1 and will be removed in v2.0.0. You can import it from pytorch_lightning.utilities instead.
rank_zero_deprecation(
Launching Web UI with arguments: --use-zluda --update-check --skip-ort
Warning: caught exception 'Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx', memory monitor disabled
Loading weights [6ce0161689] from C:\SD-Zluda\stable-diffusion-webui-amdgpu\models\Stable-diffusion\v1-5-pruned-emaonly.safetensors
Creating model from config: C:\SD-Zluda\stable-diffusion-webui-amdgpu\configs\v1-inference.yaml
C:\SD-Zluda\stable-diffusion-webui-amdgpu\venv\lib\site-packages\huggingface_hub\file_download.py:1150: FutureWarning: resume_download is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use force_download=True.
warnings.warn(

[ruby yoke]

you need to edit your sytem path. what OS are you running

[gentle lava]

i already did it

[ruby yoke]

okay

[gentle lava]

help me with one thing

[ruby yoke]

i can try

[gentle lava]

when i try to run webui this error show

Failed to load ZLUDA: Could not find module 'C:\SD-Zluda\stable-diffusion-webui-amdgpu.zluda\nvcuda.dll' (or one of its dependencies). Try using the full path with constructor syntax.
Using CPU-only torch

and

no module 'xformers'. Processing without...

[ornate elk]

Try delete the .zluda folder from the webui folder and relaunch

[buoyant palm]

Is this an error?

C:\Users\Julio\Documents\SDInstall-files\system\python\lib\site-packages\huggingface_hub\file_download.py:1150: FutureWarning: resume_download is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use force_download=True.
warnings.warn(

do i need another version of python?`i just downloaded the last version

[ruby yoke]

you need to be on 3.11.x

[buoyant palm]

yeah. Thanks. I unistalled everthing and just reinstalled it

[buoyant palm]

how do i get openpose editor for automatic 1111?

[patent fog]

anyone know how to smoothly loop a gif in AnimateDiff? Am new to stable diffusion

[vocal burrow]

not an error, it's a deprecation warning. it shouldn't matter (for now)

not sure what you're trying to install but most SD related software (apart from comfyUI and a few others) require a python in between 3.10.6 and 3.10.11. So using 3.11 might cause issues

[stiff mountain]

oops, maybe this channel was more appropriate!

I updated Stable and now it seems to me that it takes a long time to generate even simple images, furthermore the results dont resemble the model or the Lora I'm using, I fear some of my settings are wrong, could anyone help me? ;v; Im now trying to reinstall it 💧

[vocal burrow]

what do the results look like ? what settings are you using ? care to generate an image and send a full consonle log (the white text in black window thing, copy paste it in a .txt and drop said file in here)?

[vocal burrow]

you install it through the extension tab

[buoyant palm]

I installed this one: v3.10.6. thanks. I am trying to do controlled images.

[stiff mountain]

I tried to delete stable and reinstall it from github, but after have running "update" now "run" dosent works anymore 🫠

that's how my terminal ends like:
"...
File "email\message.py", line 15, in <module>
File "email\utils.py", line 29, in <module>
File "socket.py", line 51, in <module>
ModuleNotFoundError: No module named '_socket'"

[vocal burrow]

what are you installing exactly ?

[stiff mountain]

https://github.com/AUTOMATIC1111/stable-diffusion-webui/releases/tag/v1.0.0-pre this one!

[vocal burrow]

this is ancient, don't use that package please

what's your gpu ?

[stiff mountain]

Im running a 3060 ti!

[vocal burrow]

ok then follow those instructions https://github.com/CS1o/Stable-Diffusion-Info/wiki/Installation-Guides#nvidia-automatic1111-webui-stable-diffusion-webui it will install automatic1111's stable-diffusion-webui

(url is different because it's just a guide rewriting the official instructions from here https://github.com/AUTOMATIC1111/stable-diffusion-webui?tab=readme-ov-file#automatic-installation-on-windows with more words and additional nice to have for better speed.)

[twin wharf]

anyone know any command line auguments to improve speed of flux on zluda?

[ruby yoke]

you should start by reading through @ornate elk guides that are pinned in this channel

[stiff mountain]

Ok Im done!

I'll take back some model\lora and test again, thank you! 🙏

[stiff mountain]

ok that's an exemple, I put a very simple and silly promt but it still I feels like it dosent really recognize or use anyore any of my lora\models?

Also its really slow at generete even something silly like this :x

(that's the masterpiece I created btw)

[vocal burrow]

This is a pony model, therefore it's based on SDXL which is trained with 1024x1024 reesolution images. Asking for 512x512 with it is too low, you'll have to bump it up to at least 768x768 to get descent results.

Also don't use 0 for hires steps cause 0 means "use the same number of steps as for the original output" (so in your screenshot => 20)

10 to 15 should be enough

or simply don't use the hires.fix and like I said, set directly a 1024x1024 output resolution.

[stiff mountain]

!

I'll try whichever solution seems most convenient to me, but there's nothing to do about the speed? Does the model have any influence? Because before it didn't take so long :S

[ruby yoke]

you'd be better off using the base model

[vocal burrow]

1. generating straight to 1024x1024 will be faster
2. sd 1.5 based models are the fastest but output 512x512, sd2.1 output 768x768 kinda middle ground but not that much used.
3. There are "lightning" and "turbo" variant of some models designed to achieve correct results in 2 to 3 steps
4. the more lora you use the slower it will get
5. did you add --xformers --no-half-vae --medvram like stated by the guide ?
6. there might be something broken with your install, we'd need to see a full log for that and I mean the full log, from the top.

[stiff mountain]

I just put --xformers --no-half-vae, should I also add --medvram?

[vocal burrow]

3060 ti => 8gb => so at the very least --medvram-sdxl
Which means "--medvram but only for sdxl models" as sdxl models are hungry for that vram.

--medvram is a trade off, basically it will load parts of the model, do part of the work, then unload it and load the rest to finish the work.

on gpu with enough vram it's definitely slower to use it. But with """low vram gpus""" not using it will definitely harm your performance as it might use your system ram to fill the gap.

and storing gpu stuff in ram is slooooooow.

[stiff mountain]

could this be the reason for these times? o:

[vocal burrow]

could be

but hey like I said, let me some logs to be sure.

nearly 3 am here so I'll probably check out pretty soon

[stiff mountain]

Same here @@

where can I find the log??

[vocal burrow]

it's in the console you screenshoted earlier

just copy paste everything top to bottom into a .txt and drop said file in here

(only "personal information" in here should be your windows username. So probably not a concern)

[stiff mountain]

Ok, let me just restart the terminal first!
Should I gen at least one image before sending the log?

[vocal burrow]

probably not necessary from what I saw before

[stiff mountain]

venv "S:\AI\stable-diffusion-webui\venv\Scripts\Python.exe"
Python 3.10.11 (tags/v3.10.11:7d4cc5a, Apr 5 2023, 00:38:17) [MSC v.1929 64 bit (AMD64)]
Version: v1.10.1
Commit hash: 82a973c04367123ae98bd9abdf80d9eda9b910e2
Installing xformers
Launching Web UI with arguments: --xformers --no-half-vae --medvram
Tag Autocomplete: Could not locate model-keyword extension, Lora trigger word completion will be limited to those added through the extra networks menu.
Checkpoint v1-5-pruned-emaonly.safetensors not found; loading fallback ponyDiffusionV6XL_v6StartWithThisOne.safetensors [67ab2fd8ec]
Loading weights [67ab2fd8ec] from S:\AI\stable-diffusion-webui\models\Stable-diffusion\ponyDiffusionV6XL_v6StartWithThisOne.safetensors
Running on local URL: - - - -

To create a public link, set share=True in launch().
Creating model from config: S:\AI\stable-diffusion-webui\repositories\generative-models\configs\inference\sd_xl_base.yaml
S:\AI\stable-diffusion-webui\venv\lib\site-packages\huggingface_hub\file_download.py:1150: FutureWarning: resume_download is deprecated and will be removed in version 1.0.0. Downloads always resume when possible. If you want to force a new download, use force_download=True.
warnings.warn(
Startup time: 22.2s (prepare environment: 14.3s, import torch: 2.9s, import gradio: 0.9s, setup paths: 1.4s, initialize shared: 0.2s, other imports: 0.6s, load scripts: 0.7s, create ui: 0.5s, gradio launch: 0.5s).
Applying attention optimization: xformers... done.
Model loaded in 13.7s (load weights from disk: 0.8s, create model: 0.6s, apply weights to model: 11.0s, calculate empty prompt: 1.1s).

[vocal burrow]

127.0.0.1:7860 is a local ip no needs to hide it

[stiff mountain]

https://tenor.com/view/marcoelprro-gif-27525382

[vocal burrow]

ok so no red flags during the initialization.

Try it like that and see how it get

[stiff mountain]

the anathomy is kinda broken, but the lora and model looks work now and the speed is not that slow!

[vocal burrow]

how slow ?

[stiff mountain]

Its mostly the last 2% which take a while

but on terminal it already say 100% so Im not sure what does it means :S

lets say that I started an image one minute ago and now its at 60%!

[vocal burrow]

last % always take a while, it's normal.

[stiff mountain]

a batch of 4 actually

not sure about how could I fix the anatomy but I'll give a check tomorrow, rn here are the 3 am x_x

[vocal burrow]

you probably don't want to do that.

the larger the batch the more vram needed.

use batch count instead to be safe.

batch of 4 might be too much, batch of 2 should be good.

batch size <=> how many pics to generate at the same time / in parallel
batch count <=> how many pics to generate back to back.

[stiff mountain]

so its better play with batch count that batch size?

or should I just avoid both of them?

[vocal burrow]

batch count is safe to use

[stiff mountain]

((ok sorry my internet died and I didnt read in time before sending these messages))

[vocal burrow]

batch size will use more vram and can cause slow down / error if there s not enough

[stiff mountain]

I'll go to sleep for now, idk when my pc will send these message but thank you a lot for all the support!

[sonic ice]

anyone able to help? i still cant get it to make faces right

like, without loras its still messing up teh faces

[tranquil terrace]

@halcyon zodiac I have a partnership praposal

[untold hollow]

anyone have experience installing svd(sv3d) on mac m1? ive been stuck for hours

im getting
CFLAGS = None

    --- stderr
  
  
    error occurred: Failed to find tool. Is `clang` installed?
  
  
  error: `cargo rustc --lib --message-format=json-render-diagnostics --manifest-path Cargo.toml --release -v --features pyo3/extension-module --crate-type cdylib -- -C 'link-args=-undefined dynamic_lookup -Wl,-install_name,@rpath/tokenizers.cpython-310-darwin.so'` failed with code 101
  [end of output]

note: This error originates from a subprocess, and is likely not a problem with pip.
ERROR: Failed building wheel for tokenizers
Failed to build tokenizers
ERROR: ERROR: Failed to build installable wheels for some pyproject.toml based projects (tokenizers)

but pip install tokenizers works fine and i have gcc, clang installed:
(.pt2) (base) michellechen@MacBook-Pro-14-inch-2036 generative-models % gcc --version
Apple clang version 15.0.0 (clang-1500.3.9.4)
Target: arm64-apple-darwin23.5.0
Thread model: posix
InstalledDir: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin
(.pt2) (base) michellechen@MacBook-Pro-14-inch-2036 generative-models % clang --version
Apple clang version 15.0.0 (clang-1500.3.9.4)
Target: arm64-apple-darwin23.5.0
Thread model: posix
InstalledDir: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin

on Python 3.10.14

[vocal burrow]

show examples of results and settings

also that s probably a dicussion for #📝｜prompting-help

[vocal burrow]

error occurred: Failed to find tool. Is clang installed?
Welp seems like you need to install clang

[untold hollow]

clang is installed tho?
Apple clang version 15.0.0 (clang-1500.3.9.4)
Target: arm64-apple-darwin23.5.0
Thread model: posix
InstalledDir: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin

[vocal burrow]

I don t have the full log here so I can t help much. Especially considering that the last apple machine I owned was probably some apple 2 grey box.
But is clang installed in your crate/container/venv/whatever may apply there ?

[untold hollow]

its installed here: /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/clang

[untold hollow]

should it not be in xcode?

[vocal burrow]

That s probably (one of) its proper place system wide. But
1/ maybe what you re trying to use or install ( don t even know ) is running in some sort of vm/venv/cargo where it s not defined
2/ /Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/clang might not added to your path, or some other magic env variable might not set and therefore clang cannot be found.

[severe kindle]

@untold hollow I'm guessing svd is trying to install tokenizers 0.12.x and you're on Apple Silicon. The 0.12 series was never released as a binary wheel for Apple Silicon, which is why its trying to build it from source.
It's probably wanting that version due to an outdated (as in 2 years old) version of transformers so I'd try updating transformers to a version that worked with tokenizers 0.13.x.
Alternatively you could try downgrading tokenizers to 11.6

[rugged latch]

imagine being the one peson who comes here, and asks, "what is stable diffusion "

[round ravine]

Can anyone help, i'm stuck at this on first run

[vocal burrow]

it s normal for zluda stuff to take A WHILE for their first run

talking 10 to 40 minutes depending of the hardware.

[round ravine]

does this apply for all models?

[vocal burrow]

It should be good after that once it s done "translating cuda stuff to amd"

[vocal burrow]

doesn t matter the model you loaded

it will be slow for the first run and should be good afterwards regardless of the model.

[round ravine]

ok let me wait and come back later then

thank you

[burnt vortex]

Do I come here for problems withthe website?

[vocal burrow]

Probably not, this tech support is mainly for people hosting SD instances locally. I d advise you to use the website's support page.

[grim moon]

which one should i download?

Im installing automatic111 with zluda

[ornate elk]

What's your GPU?

[grim moon]

Oh hi CS1o

[grim moon]

Rx 6900XT

[muted scarab]

I saw a video saying that you should have 8 GB or more of dedicated ram?* for SDXL
I have 16GB of RAM, how can i get more VRAM?

[errant thistle]

So im using automatic 1111, and when i look at my lora tab, it only shows 18 of my loras.
Is it only able to display a certain amount before it caps?

Or is there just a bug where some loras just dont appear? Because ive had it with a few even before i downloaded that many.

[muted scarab]

there's a setting to show lora that aren't compatible with your SD vers, there's also a disable description on lora panels setting

[errant thistle]

Uhuh. Where abouts is the first setting?

[muted scarab]

Settings / Extra Networks / ^ Always show all networks on the Lora page

[errant thistle]

Yeah that brought them back, all 9. Dang. So i assume the ones that were otherwise hidden just still wouldnt work with automatic 11111 even if i used them?

[muted scarab]

idk i did that an ran it and i froze my pc and i had to restart, i am noob at this

so im downloading matching sd vers and models now

it says in the terminal what SD vers you are using, i think it's near the top

[errant thistle]

"Version: v1.10.1"

So its not necessarily incompatible with sd automatic111 in general? Its just the version im using that doesnt like it?