🤝｜tech-support

[haughty zodiac]

The result is catastrophic by the way :c

the lora wasn't trained enough but each step being 20s makes it so long

[unique bluff]

I was told I can’t use stable diffusion on my computer because it uses the Intel Iris Pro Graphics 62000 1536 MB as GPU. Would that be correct?

[low tree]

Hi ! i need help :3
I use stable diffusion locally and I would like to add the zaxychromaXL module in order to have access to SDXL Styles and ControlNet Integrated.
I put zaxychromaxl_v60.safetensors in models > stable-diffusion but when on stable diffusion I put this models, nothing happens..

[severe sand]

The only thing I changed are the starting parameters as per @ornate elk's suggestions

i also tried Euler and a different model before, same results

CS1o said that it's most likely a ZLUDA problem

https://discord.com/channels/1002292111942635562/1229200167782322247

posted some more info there

[nova wren]

anyone knows if its possible to create a batch job with a word with different values for each run?

i want to create a few hundred pictures but i need to use different colors or parameters for each generation in a cyclic way

kind of like an array...

[short birch]

Complete neophyte here, so apologies if I get confued easily. I have been using A1111 on my AMD machine by using PyTorch 2.0.0+cpu just fine for a few weeks, but this morning it updated and now I get a long list of exceptions then the concolse window closes out. Does anyone have any idea what I need to do to get it running again?

[nova wren]

--reinstall-xformers on your webuiuser.bat file

i have had that break quite a few times... but usually you can just force reinstall and then once its working good you remove the flag and start sd again

[true kettle]

I need a help in getting the cartoon style has target of input images

[nova wren]

ah ignore my last question solved it already

@true kettle you trying to turn into a cartoon the photo? also probably for the best not to post pictures of real people...

[vocal burrow]

I would add --reinstall-torch --reinstall-xformers to your command line arguments as it seems like your whole pytorch is outdated. (Weird that it s not telling you to run those tho.) Run it once so it actuall reinstall torch and xformers and then remove the arguments so it doesnt reinstall everything everytime you launch it

[true kettle]

@nova wren Its the person from publicly available dataset I just want the facial features in the cartoon of input image

I want a image to image translation model but the main constraint is data

[nova wren]

usually for that kind of situation i use inpainting with depth preprocessor from controlnet... works wonders

[short birch]

Thank you so much for the suggestion! I added the changes and ran it once, but xformers still seems to have a problem. When I first started using SD i had to get a special version of PyTorch that uses AMD firmware, and never had access to things like inpainting or training due to this.

[true kettle]

@nova wren can you share the repo and guide me how to achieve that

[nova wren]

@true kettle not sure if there is a guide out there... for that specific flow, but let me see if there is one, if not maybe i can send you a few screenshots on how to set it up in a bit

[true kettle]

thanks @nova wren

[vocal burrow]

wait... are you trying to use xformers with an AMD gpu ?

[nova wren]

i think he might have been... i kind of glossed over that.... anyway has anyone figured a way to get tensorRT to work with control net?

[severe sand]

@vocal burrow ?

[vocal burrow]

looks ok, what does the log says when you try to generate it ?

[severe sand]

idk why there are 4 progress bars

with 1 at 80%

[vocal burrow]

sorry I don't know, it looks like a zluda specific error. I don't have an AMD gpu to replicate it

[severe sand]

Alright :((

can you spot anything "off" in the console?

[vocal burrow]

apart from the startup time taking forever, not much

[valid lichen]

Anyone know why the defaul lineart model isn't available

I'm on colab and I'm not sure where I'm suppose to check this

It shows up on the X/Y/Z plot

Also if I run lineart_realistic vs lineart it gives me different result :/

[drifting kiln]

does anyone have any experience using 3d openpose?

idk what preprocessor i should use

or what model

it just doesnt work

[dense viper]

How can I get access to the #platform channel?

[dreamy lodge]

RuntimeError: CUDA error: out of memory
CUDA kernel errors might be asynchronously reported at some other API call, so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1.
Compile with TORCH_USE_CUDA_DSA to enable device-side assertions.

Any idea how to fix this?

[fallen quest]

ComfyUi?

[dreamy lodge]

a1111

[fallen quest]

Ok, sorry. No idea

[limber pier]

sounds like you're out of vram

a1111 isn't very eofficient

[ornate elk]

What were your image settings?

[dreamy lodge]

dpm karas 2++, 512x768, latent hires 2x, what else?

happens after a few generations

[surreal flame]

/dr

[ornate elk]

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

[dreamy lodge]

nvidia rtx 3050 laptop. its the settings u told me last time

[ornate elk]

Okay, so --xformers --medvram --no-half-vae

What's the model? Sdxl or 1.5 ?

[dreamy lodge]

1.5

[ornate elk]

Okay then you need the tiled VAE extension.
It can prevent that out of vram when using hires fix

Also make sure to not run anything in the background. Like games or wallpaper engine

[dreamy lodge]

"TiledDiffusion with Tiled VAE
manipulations" this?

[ornate elk]

Yes

And then only enabled the Tiled Vae and leave the settings on default

[gilded canyon]

Hey, I'm trying to load in an SDXL model in WebUi Forge, but I am getting a connection error when I select it and this traceback

[ornate elk]

Your missing a VAE for that model.
That fixes such spots and color issues
More info here:
#🤝｜tech-support message

[ornate elk]

Yes it won't work with that GPU chip. But it would work on the very slow CPU mode.

[ornate elk]

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

[ornate elk]

Hey, what's your GPU?

[ornate elk]

That's doable with the Controlnet and the IP-Adapter Full Face model

[ornate elk]

Make sure the webui is whitelisted in any browser adblocker

Also can you try it in auto1111 again?

[gilded canyon]

okay sure

[ornate elk]

But before make sure you have --xformers --no-half-vae in the webui-user.bat

At the line Commandline_Args=

[gilded canyon]

okay ill try that right now

Ok so, this is off topic but I deleted venv because i recently upgraded my GPU and wanted to reinstall stuff for it, but now im getting a MemoryError when python tries installing torch and torchvision

[ornate elk]

Can you screenshot the error?

[gilded canyon]

I have 16gb of ram and on my old GPU it installed fine ;-;

[ornate elk]

You should adjust the windows page file for C

[gilded canyon]

how do i change that

[ornate elk]

Here is a guide:
Make sure you have 15-20gb free space on C:/

Set the pagefile size to 16000min and 24000 Max
For C. Make sure its disabled for all other drives.
Then reboot the PC. (Important)

How to adjust the Pagefile Guide: https://mcci.com/support/guides/how-to-change-the-windows-pagefile-size/

[gilded canyon]

I did this, now im getting this error

[ornate elk]

No space left on device

[gilded canyon]

Do i have to install on C: or can I install on D:

my windows installation takes up a lot of space on my C: since thats all I use it for (SSD)

[ornate elk]

Install what?

[gilded canyon]

torch

[ornate elk]

Page file or SD?

Is the other drive a HDD?

[gilded canyon]

yes

I have the SD folder on my HDD, but when I run it, the venv folder installation just installs on my C

im pretty sure at least

cus idk how i have no space left

[ornate elk]

Okay, then you can do the following:
Move your models/stable-diffusion folder onto the HDD.
Edit the webui-user.bat and there add --ckpt-dir "D:/Path/to/theNewModels/" folder location

You have to set the right path to it. Mine is just a placeholder

[ornate elk]

Wait what

So you have SD on the HDD?

Then you don't need to move the folders

[gilded canyon]

Yes, I just dont have space for the stuff Python installs cause that installs on my C drive

like Torch and Torchvision and stuff

does that make sense

[ornate elk]

Yes it Caches on C, and that can't be changed.

[ornate elk]

Download the Wiztree tool and check what's taking up your space.
Clean the dustbin too.

[gilded canyon]

Okay got it all working, now WebUI launched

I'm running the arguments you told me before (xformers and no half vae), and when I select my SDXL model, I get this traceback:

changing setting sd_model_checkpoint to autismmixSDXL_autismmixPony.safetensors: AttributeError Traceback (most recent call last): File "D:\Downloads\StableDiffusion\webui\modules\options.py", line 165, in set option.onchange() File "D:\Downloads\StableDiffusion\webui\modules\call_queue.py", line 13, in f res = func(*args, **kwargs) File "D:\Downloads\StableDiffusion\webui\modules\initialize_util.py", line 181, in <lambda> shared.opts.onchange("sd_model_checkpoint", wrap_queued_call(lambda: sd_models.reload_model_weights()), call=False) File "D:\Downloads\StableDiffusion\webui\modules\sd_models.py", line 860, in reload_model_weights sd_model = reuse_model_from_already_loaded(sd_model, checkpoint_info, timer) File "D:\Downloads\StableDiffusion\webui\modules\sd_models.py", line 793, in reuse_model_from_already_loaded send_model_to_cpu(sd_model) File "D:\Downloads\StableDiffusion\webui\modules\sd_models.py", line 662, in send_model_to_cpu if m.lowvram: AttributeError: 'NoneType' object has no attribute 'lowvram'

oh nvm

i tried again and it worked

let me try generating

[eager halo]

👋🏻 When I use the API, it doesn't use any credits/tokens in my dashboard. Not really complaining if its giving me free use, but if I create images in the dashboard, it does use them. Is this expected behavior?

[vapid patrol]

Hello ,
Please guide me to install Forge SD on amd card

[dreamy rampart]

Ok so I have A1111 since last year but I am an idiot with this stuff

So I have a few models in the model folder and 1 random YAML.

Should I be renaming the yaml or something?

I feel like a caveman

[tight scroll]

Can someone help me? I'm trying to make the new PAG node work for ComfyUI. But it is not appearing in my node list due to this error. "0.1 seconds (IMPORT FAILED): C:\A.I\ComfyUI\ComfyUI_windows_portable\ComfyUI\custom_nodes\sd-perturbed-attention". the first isue is that it's telling me that the samplers.py does not exist when it clearly does. As you can see it's inside C:\A.I\ComfyUI\ComfyUI_windows_portable\ComfyUI\comfy. Second it's this module called 'ldm_Patched' is missing, which I have no idea what that is. -- SOLVED

[ornate elk]

Forge doesn't support AMD cards that well. You better use one of my two guides for AMD with auto1111.
Both guides are in the pinned messages if this channel.
The zluda guide works for 5600xt, 5700xt, 6 and 7000 series.
If you have an older GPU then you need to use the directml guide.

[ornate elk]

A yaml is mostly for a model that needs it. It should have the same name as the model to get loaded correctly

[woven locust]

Can someone help me with a prompt template for garment generation with specific fixed attributes required in Fashion Garment industry.

[normal silo]

[loud marlin]

Hello. So the guide you provided allowed me to at least install the directories. I now get RuntimeError: Torch is not able to use GPU. I added the --skip-torch-cua-test to COMMANDLINE_ARGS and it didn't resolve the issue. I did some of my own research and found out that nVidia doesn't have official driver support for their laptop GPU (laptop is 2 years old). I also noticed that the laptop shows an onboard graphics card as well and am wondering if it's not selecting the nVidia when i run the .bat file? Are there more commands in the DOS to explain why Torch would not be able to use GPU? Sorry for the long post but wanted to show I am actively trying to get it resolved on my own. Thanks.

[ornate elk]

Hey, what's your nvidia GPU?

[loud marlin]

nVidia GeForce RTX 3070 Laptop GPU

[ornate elk]

This GPU should have a supported driver by nvidia

Select the right version here:
https://www.nvidia.com/Download/index.aspx

[loud marlin]

Would you recommend studio driver be preferred over game ready driver? I was able to find the right version. I was there before but didn't see it. Thanks!

Also it just shows Windows 11 available but not 64-bit version? Still safe to run I would presume but want clarification.

[karmic nova]

has anyone run into this issue where red dots appear? It does it regardless of photo source or cfg or anything

[vocal burrow]

Go with the game ready driver.

[vocal burrow]

Windows 11 32 bit does not exist

[vocal burrow]

try with a different vae

[karmic nova]

no vae used, just a model in auto1111

[noble belfry]

if you run a xyz for each step in the process you can see the dots appear out of the noise very early on. its an old SD1.5 issue

also, try it with a vae. 🙂

[ornate elk]

Use the gaming driver 64 bit

[loud marlin]

so Windows 10 ok?

[ornate elk]

Choose windows 11 if your on windows 11

And 10 64 bit if your on 10

[loud marlin]

Sorry for the confusion but I have Windows 11 64-bit and only Windows 10 64-bit available. I googled if it would be ok and it seems to be the case as it's the same architecture (the only other option was Windows 11. Gonna give it a whirl.

[ornate elk]

Yes windows 11 is only for 64bit

[ornate elk]

You need to use a vae

Then the dots will disappear

[fair oxide]

@ornate elk have you tried this sampler https://github.com/licyk/advanced_euler_sampler_extension

[ornate elk]

Hey, nope not yet, is it good?

[fair oxide]

i tried out after reading the reviews but so far what i've noticed the background with this sampler comes out slightly blurred

not entirely sure if im overlooking anything, but tried it with 30 steps, 7 cfg

btw @ornate elk there's been update on a1111

this was my xyz result #🤝｜tech-support message

[viscid badge]

hello

when I try to render with controlnet setting I have an error message :

" TypeError: 'NoneType' object is not iterable "

I dont know wha to do, i show u my settings maybe I made a mistake..

[fair oxide]

where did you download the model from?

[loud marlin]

I'm in. Thanks for your tremendous help. Appreciate your knowledge and guidance!

[ornate elk]

nice, no problem 🙂

[woeful turret]

What GPU would you recommend for machine learning, ai etc. I have been looking at RTX 3060 12GB. Any other suggestions in that price range?

[vocal burrow]

3060 12gb is probably still the best bang for the buck (when it comes to AI)

[civic mirage]

could someone help me get ponyxl to work? i cant find any guides

[short niche]

help!
I'm trying to use open pose, but I cant upload the img, the "box" is not there in control net ... what can be the reason?

[rugged summit]

can someone explain to me what the difference between Model Lora and VAE is in non complicated terms

[ornate elk]

your in img2img, so it will use the box of img2img
in your screenshotr you can also check the button "Upload independent control image" to get the controlnet box back

[ornate elk]

Model (checkpoint) = needed to generate images. can be trained on different styles, and merged with other models. (file sizes mostly from 2-6gb)
Most models are based on SD Version 1.5 or SDXL
Lora = small files that are trained on specific stuff like a Style or Character, that goes on top of a model to help it generate the right Character/style.
Vae = needed for color correction for different models. If you output image is not vibrant and more pale or grayed out, you need to use a vae with that model
Some models come with a custom Vae already included

[short niche]

thank you very much ^^

[rugged summit]

can i use a model with an included vae with a different vae

[ornate elk]

yes, that will use the not included vae then

[rugged summit]

thank you

[ornate elk]

no problem 🙂

[glad fiber]

what is the best model to convert a pixelated art to a realistic?

[proud jay]

Hey, can I get some help?
Here's my set up:
AMD 6600
AMD Radeon (TM) Graphics
Windows 10

I realize that SD is not really meant for Windows much less AMD gpus, but I'm not trying to use the other features that Linux and Nividia users have. All I want is for SD to use my 6600 instead of my CPU

[glad fiber]

#🤝｜tech-support message

[ornate elk]

hey, a 6600 is supported by my AMD Guide with zluda, its in the pinned messages of this channel

[proud jay]

gracias

[rugged summit]

is it normal that a gtx1650 4gb with 16gb of ram takes about 10 minutes to generate a 256x540 picture?

it's not really a big deal at least it works but is this the average time it takes for this gpu?

[ornate elk]

I edited the Guide right now, someone made a tweak so it will work even faster on your GPU

[ornate elk]

No, whats in your webui-user.bat?
also what model did you used?

[proud jay]

Hey, since I have a 6600, I assume that you want me to rename the Optimizised_ROCmLibs_gfx.7z extract to "library" to make sure things line up, yeah?

[rugged summit]

set COMMANDLINE_ARGS=--xformers --medvram --no-half --precision full --upcast-sampling
set PYTORCH_CUDA_ALLOC_CONF=garbage_collection_threshold:0.9,max_split_size_mb:512
these are the only things i changed, when i first started it was the same time just full of bugs and most of the times didnt generate, im using sd1.5 models

[ornate elk]

make a new folder and name it library, then extract the files in there

the other library folder needs to be renamed before that

[proud jay]

oh so the library is nested in another folder named library?

I just moved it over and renamed the folder

[ornate elk]

ah alright, that works, if you now have two folders (library and old_library) then its okay

just make sure in library folder isnt another folder, onle the raw files

[proud jay]

also, just to be clear with the PATH instructions, you want this to be its own path?

[ornate elk]

yep 👍

[ornate elk]

remove --precision full --upcast-sampling
and remove set PYTORCH_CUDA_ALLOC_CONF=garbage_collection_threshold:0.9,max_split_size_mb:512
then saveand relaunch the webui-user.bat

also make sure your using an model with a size of 2gb

[rugged summit]

why's that?

i mainly use 2gb ones but i also used some 4gbs

[ornate elk]

you can use 4gb ones too, but 2gb ones will work faster

i wouldnt use sdxl models with that card as its really slow

[ornate elk]

I just said that because some people load Lora's as models and then get issues like black images or freezing webui

[rugged summit]

i have an issue some times where the images is fine in the preview then goes totally gray, it works if i swap the model and then put it again

im generating a picture rn and it's going at abt 15.56s/it

it was generating fairly fast at the beginning and it told me ETA 3 mins, then at 2:50 it stopped and it started going up and slowing down a lot

[ornate elk]

Is your driver updated?

Make sure to not run any heavy programs like Wallpaper Engine or games in the background

[rugged summit]

nope seems like i havent updated them in a while

[rugged summit]

wallpaper engine shouldnt take that many resources

i check the usage and sd doesnt seem to use all my ram and theres much space left, only thing it's using is my gpu

[ornate elk]

Wallpaper Engine causes massive issues when using SD. It even can crash the system xD

[rugged summit]

holy i didnt know that

alright im going to update drivers and turn of wallpaper engine rq

[ornate elk]

Perfect. Make a reboot after the driver update and Don't "pause" WP Engine instead close/shut it down when using SD

[proud jay]

aight, almost there

[ornate elk]

Oh that is a new error

[ornate elk]

What zluda version did you downloaded?

[proud jay]

3.7

since 3.8 is a pre-release version i figured it might be more unstable

I figure you have something coming up but this is where that error is being raised

the bottom one specifically

[ornate elk]

Is your zluda folder correctly added as path in the system variables?

[proud jay]

It's gone... tf?

Like its actually gone

okay

[ornate elk]

Did it worked?

[rugged summit]

i just restarted imma try it rn

[proud jay]

readded the paths

almost there

wait,

it can't be

[glad fiber]

how can i get the controlnet's image on comfyui? like, i want the depth image

[ornate elk]

Do you have a ryzen CPU with igpu?

[proud jay]

sort of

the 6600 should be its own discrete gpu though

[ornate elk]

Yes, check the device manager and there under Graphicadapter if you have two entry's there

[proud jay]

pretty sure the AMD Radeon (TM) Graphics is the integrated one

[ornate elk]

Yes, that need to be disabled in device manager

Zluda detected it as primary gpu

[proud jay]

man

[ornate elk]

Need to go off as its pretty late

Let me know if you got it to work or need help.

[hollow schooner]

Ello anyone know what’s with mine started yesterday whenever i try to generate it goes to like 49% and won’t go any further

Says error connection timed out

Amd Radeon RX 6600

[rugged summit]

using 4gb model, DPM++ 2M KARRAS, 576x768, around 10-11-12s/it

am i gonna get faster generation with 2gb model? maybe 5s/it?

[hollow schooner]

the webui-user settings

[rugged summit]

also this is the load time, im using a mechanical drive, am i gonna get faster speed by using my ssd or is it still gpu based

took 150s to start generating

[jolly gorge]

is there a way to make a previously generated image the preview on a model card in automatic?

[vocal burrow]

2gb is definitely --lowvram territory not --medvram especially if you're using AMD and directml

consider zluda for better performances (cf pinned comment)

[rugged summit]

im on nvidia

[vocal burrow]

Also that's the time you're wasting because of your disk

[rugged summit]

okay ill see if i can fit everything on my ssd

[vocal burrow]

then why are you using directml ?

[rugged summit]

can i just copy paste the sd folder

[vocal burrow]

why are you all those command line args

[glad fiber]

how can i get the controlnet's image on comfyui? like, i want the depth image of my input image

[rugged summit]

idk bro i got them from reddit when i was trying to solve some bugs

[vocal burrow]

also that's NOT how to use set VENV_DIR line

[rugged summit]

i DO NOT know how any of this works 😭

[vocal burrow]

tbh I'd say throw whatever you have installed at the moment in the garbage bin and start fresh following the pinned guide for nvidia then

[vocal burrow]

Preview syntax rules, put .png next to the [embedding/model/etc] and name as such :

• Embedding : [embedding].pt -> [embedding].preview.png
• LoRa : [lora].safetensors -> [lora].png
• Models : [Model].ckpt/.safetensors -> [Model].png

[rugged summit]

i just checked the guide and it is exactly what i did

[vocal burrow]

oh wait I confused soyzree screenshot for yours

ok then what's in your webui-user.bat ?

[rugged summit]

set COMMANDLINE_ARGS=--xformers --medvram --no-half

[vocal burrow]

set COMMANDLINE_ARGS=--xformers --lowvram

try that

[rugged summit]

100%|██████████████████████████████████████████████████████████████████████████████████| 20/20 [03:20<00:00, 10.02s/it]
tiled upscale: 100%|███████████████████████████████████████████████████████████████████| 20/20 [00:03<00:00, 5.26it/s]
100%|██████████████████████████████████████████████████████████████████████████████████| 12/12 [17:54<00:00, 89.53s/it]
Total progress: 100%|██████████████████████████████████████████████████████████████████| 32/32 [23:13<00:00, 43.56s/it]
{"prompt": "((masterpiece, best quality)), close-up, straight on, face focus,((1 girl)), {long hair}, ((big breasts)), standing, hair ornamen, gloves, (incredible absurdity), gilded frame, (detailed light),lighting, colorful, layered backgrounds, (gothic architecture), pond, (gorgeous background), dynamic angle, delicate background, (shine),ganyu, twin horn, bare shoulder, bodysuit lora:ganyu2:0.65", "all_prompts": ["((masterpiece, best quality)), close-up, straight on, face focus,((1 girl)), {long hair}, ((big breasts)), standing, hair ornamen, gloves, (incredible absurdity), gilded frame, (detailed light),lighting, colorful, layered backgrounds, (gothic architecture), pond, (gorgeous background), dynamic angle, delicate background, (shine),ganyu, twin horn, bare shoulder, bodysuit lora:ganyu2:0.65"], "negative_prompt": "EasyNegative, extra fingers, fewer fingers, (worst quality, low quality:1.4), loli,nsfw,text", "all_negative_prompts": ["EasyNegative, extra fingers, fewer fingers, (worst quality, low quality:1.4), loli,nsfw,text"], "seed": 2971847726, "all_seeds": [2971847726], "subseed": 3668259256, "all_subseeds": [3668259256], "subseed_strength": 0, "width": 768, "height": 576, "sampler_name": "DPM++ 2M", "cfg_scale": 4, "steps": 20, "batch_size": 1, "restore_faces": false, "face_restoration_model": null, "sd_model_name": "counterfeitV30_25", "sd_model_hash": "a074b8864e", "sd_vae_name": null, "sd_vae_hash": null, "seed_resize_from_w": -1, "seed_resize_from_h": -1, "denoising_strength": 0.6, "extra_generation_params": {"Denoising strength": 0.6, "Hires prompt": null, "Hires negative prompt": null, "Hires schedule type": null, "Hires upscale": 2, "Hires steps": 12, "Hires upscaler": "R-ESRGAN 4x+ Anime6B", "Lora hashes": "ganyu2: e39e31039bc3", "Schedule type": "Karras"}, "index_of_first_image": 0, "infotexts": ["((masterpiece, best quality)), close-up, straight on, face focus,((1 girl)), {long hair}, ((big breasts)), standing, hair ornamen, gloves, (incredible absurdity), gilded frame, (detailed light),lighting, colorful, layered backgrounds, (gothic architecture), pond, (gorgeous background), dynamic angle, delicate background, (shine),ganyu, twin horn, bare shoulder, bodysuit lora:ganyu2:0.65\nNegative prompt: EasyNegative, extra fingers, fewer fingers, (worst quality, low quality:1.4), loli,nsfw,text\nSteps: 20, Sampler: DPM++ 2M, Schedule type: Karras, CFG scale: 4, Seed: 2971847726, Size: 768x576, Model hash: a074b8864e, Model: counterfeitV30_25, Denoising strength: 0.6, Clip skip: 2, Hires upscale: 2, Hires steps: 12, Hires upscaler: R-ESRGAN 4x+ Anime6B, Lora hashes: "ganyu2: e39e31039bc3", Version: v1.9.0"], "styles": [], "job_timestamp": "20240417014328", "clip_skip": 2, "is_using_inpainting_conditioning": false, "version": "v1.9.0"}

this is a gen i just did, about 3 minutes for the image but it took 20 minutes for upscaling

size 768x576 upscaled 2x latent 12 steps 0.60 strenght

[vocal burrow]

what's your gpu again ?

[rugged summit]

is it normal that the upscaling is taking so long

gtx1650

4gb

[vocal burrow]

oh dang

so yup

you had the correct command line args then

[rugged summit]

i usually do 256x512 then upscale, right now im doing wallpapers so im doing them at abt double the quality

also 4gb model, it should take way less if i do my normal pics but the upscaling is insanely slow

just checked with all i did and speed is about the same both withlatent and r-esrgan x4 but latent is way better for what im doing it seems, almost uncomparable

[vocal burrow]

1650 is pretty bad.... it's definitely an old budget gpu at this point

You're not gonna get good perfs out of it when trying to upscale.

[whole osprey]

Images are seed-like in their expression

Complexity iterates

[vocal burrow]

You're using loras so it's gonna slow down the process + you're outputting at 512*768 meaning it's gonna take more time than for a 512x512 + DPM++ 2M tends to be slower than some other sampler such as Euler A + your prompt is quite complex.

make sure you're not using your gpu for anything else when generating (playing games, youtube video, wallpaper engine, etc) make sure your drivers are updated

check the power performance profile if you're on a laptop

For now stick to "easy prompts" such as "cat" and output at native 512x512 without upscaling and see how fast it goes.

It will make your life easier trying to figure out what's wrong when working on easy prompt rather than when working with such complex prompts.

And for upscaling I'd recommend using the "ultimate SD upscale" script https://github.com/Coyote-A/ultimate-upscale-for-automatic1111 rather than the native one. Especially with low config like yours

One last thing before I go to sleep, you might want to check sd forge which is a "fork" of auto1111, you might get better speed out of it too. https://github.com/lllyasviel/stable-diffusion-webui-forge

good luck

That's all I can think off the top of my head at 2am

hope it helps, let me know which parts sounds alien to you and I'll answer when I can

[rugged summit]

ill try to figure things little by little thanks

[glad fiber]

@rugged summit i also have an gtx 1650 and i use comfyui

[rugged summit]

how is it so far?

i heard about it

[glad fiber]

this prompt took 1 minute to generate

of course, as i dont have your loras, it didnt generate the same image

[rugged summit]

upscaling and how is the wuality?

what model did you use

would you mind sending the image so i can compare it

[glad fiber]

one called anime ;-;

but from what i know about sd, the model doesnt seem to interfere a lot on the time taken to generate an image

[rugged summit]

i heard about comfyui but it looks so complicated compared to automatic1111 which im using

i should watch a lot of guides

[glad fiber]

it is complicated but it is cool

the a1111 on my pc takes 1 minute to generate a 512x512 20 steps image

while i can get that in 15 seconds on comfyui

i get arount 1s/it on comfyui and 3s/it on a1111

[rugged summit]

this is the image that i generated with that prompt

can it compare to the comfyui ones you can make with the gtx1650?

[glad fiber]

alright

finishing one here

tried to generate the same image on a1111, but i got the out of memory error on --medvram

double the time

here is an analogy of comfyui and a1111 stable-diffusion-webui's:
You can use a site builder (a1111) to create a store for your products. It is easy and convenient, but you have some limitations. If, in the future, you need to add some logic to your store, you will need to actually program it (comfyui). While programming is not easy, it gives you all the freedom you want

[hallow marsh]

Hello,
I could do with some advice on getting a research project up and running with SD - using it with Dreambooth or similar.

(I don't have much experience with using git or notebooks, though have used SD last year on a local Windows machine using some installation guides I found online.)

Goal of research:
I have 32 detailed pencil drawings I made in the forest - I'd like to use these to finesse a model and then use prompts + the tag from my images to create new forest images, and see how close to my art style the results can be. -

Stretch goal:
If possible, I'd like to structure the outputs somehow, denoting areas that might be a 'tree trunk' a 'rock' or 'creek bed'.

So far:
I've taken up a subscription to Google Colab pro - (a couple of weeks in so far) and I've tried to use EasyMode_Stable_Diffusion.ipynb - after initially trying to complete the missing steps in the process detailed in this video: https://youtu.be/9nniwr-7Em8?si=ZZzIwRGYzk3UuSLV&t=870

I'm Stuck:
I've tried to troubleshoot errors coming up in the above two methods but have started to wonder if both of these methods are somewhat outdated by how they use SD - given that the errors are either not well documented online or they seem to point towards solutions that roll back various elements to older versions to be compatible with the code. - I'm not beholden to Colab if there's better solutions but as indicated above, I'd like to do some things and have a few ways of tweaking them.

What's a good recommendation of how to proceed with this? - As mentioned, I'm not very resourceful beyond typing in solutions to errors I find, so if there's a recent method that's up to date, and has some very user friendly documentation on what errors might thrown at me I'm really keen to learn.

[wooden blaze]

Hey guys is there any stable diffusion api that allows me to do an image multi prompt?

[ornate elk]

Hey you should use the directml webui with Zluda. It will be much better for your GPU.
The guide is in the pinned messages of this channel

[hollow schooner]

Oki thanks

[hollow schooner]

Uhm what does it mean by make a new folder in my drive?

[dreamy rampart]

Ok. Thanks. I'll just leave it alone

[hollow schooner]

Stuck on step three getting zluda to run

I don’t understand how to add the stuff

[hollow schooner]

okay so followed all the steps and tried to launch it and got this

[trail sigil]

hi im new here, I need help with fixing my stable diffusion, I updated my stable diffusion just now and it is not working, it is telling me "ImportError: cannot import name 'Undefined' from 'pydantic.fields' (C:\Git\Ai\stable-diffusion-webui\venv\lib\site-packages\pydantic\fields.py)" i'm not sure what it is asking me

[ornate elk]

What's your python version?

[ornate elk]

Simply create a new folder

[hollow schooner]

Where do i check that?

[ornate elk]

Hey, delete the venv folder and relaunch the webui-user.bat

[ornate elk]

Open up a CMD and run
Python --version

[hollow schooner]

uhhhh

Python : The term 'Python' is not recognized as the name of a cmdlet, function, script file, or operable program.
Check the spelling of the name, or if a path was included, verify that the path is correct and try again.
At line:1 char:1

• Python --version

•   + CategoryInfo          : ObjectNotFound: (Python:String) [], CommandNotFoundException
    + FullyQualifiedErrorId : CommandNotFoundException

[hollow schooner]

this what it shows when i type that in

Also should i delete my other stable diffusion folder

Like my older one

also here are few things im not sure i did right

[ornate elk]

That's not right

[ornate elk]

You have to run the python 3.10.11 installer again, then click modify, next, then check the box "add python to system variables"

Then edit the Path setting again

Click on the PATH in system variables and click edit.
Then make a screenshot please then I can check

[hollow schooner]

also am i supposed to have two library folders?

one that say old and one that just library

[ornate elk]

Yes, one has to be named old_library

[hollow schooner]

okay thanks

what do i press here?

[ornate elk]

Upgrade now

[hollow schooner]

says setup was succesful

redo setup again and see if i can add to varibles?

[ornate elk]

Yes

[hollow schooner]

says suuccesful

should i delete that old path?

[ornate elk]

Make a bigger screenshot first

So we can check where we ware in the variables

[hollow schooner]

damn still forgot one name thing

[ornate elk]

Okay yes you can delete that path

Because your in the wrong section

[hollow schooner]

oki deleted

[ornate elk]

In the guide it says system variables.
And you added it in user variables

[ornate elk]

So in the screenshot at the bottom you see Path. Click on it and click edit

[hollow schooner]

oh i didnt understand the wikihow so i just watched the video below it

the gray highlighted?

[ornate elk]

Click edit and now you can ADD 2 more lines there

[ornate elk]

"New"

[ornate elk]

No don't add both in one line

You need 2 new entry's

One is for C:\ZLUDA

And the other one for the %Hip

[hollow schooner]

should i restart my pc?

[ornate elk]

Yes

[hollow schooner]

oki

And then try to open the webui user thing?

Run i mean

[ornate elk]

Yep relaunch the webui-user.bat

[hollow schooner]

Uhhhhhh

im pretty sure its installed

[ornate elk]

Delete the venv folder and relaunch

If you have one

[hollow schooner]

deleted still getting same error

[vocal burrow]

(I don't see python in their path, did you set the PYTHON var in the .bat ?)

[ornate elk]

Oh true, python wasn't added to path

[ornate elk]

But he even ticked the box

[hollow schooner]

this?

[ornate elk]

Open up a CMD and type
Where python

You should get a path

[hollow schooner]

mmmmm

[ornate elk]

In a CMD not in a powershell

[hollow schooner]

INFO: Could not find files for the given pattern(s).

[heady pine]

@hollow schooner Do this - copy path and past it next to PYTHON= and then save

Can someone explain to me the difference between OpenPose Full and DW OpenPose Full? The second one seems to be better in some cases.

Is there a download for DW OpenPose Face and Hands, or is DW OpenPose Full the only one?

[ornate elk]

DWpose is a alternative to the normal openpose.
DWpose was made by a research team and they only published the dwpose full. So no hand or face only available

[heady pine]

Alright, thanks @ornate elk

[hollow schooner]

where do i find this

[ornate elk]

Can you uninstall (not delete) every python you found and then reinstall the 3.10.11 ?

[vocal burrow]

that's what cs1o is trying to figure out by making yourun where python

[hollow schooner]

delete these both?

[vocal burrow]

By default it should install in C:\Users\username\AppData\Local\Programs\Python (or something close)

[hollow schooner]

only things that show up when i search "py"

[ornate elk]

Yea uninstall them then reinstall 3.10.11 and make sure you check the box "add python to path"

[hollow schooner]

oki

[heady pine]

The best way to download Automatic1111, or any other AI art generator, is to have all the folders in one place. It's much easier to set everything up, and it's also easy to try and fix problems when they pop up. I would say the best thing you could do right now is uninstall everything and reinstall it so everything looks like this. It's your preference, though. @hollow schooner

[hollow schooner]

uhhhhhh

[heady pine]

If you are having any problems - this is the best guide https://www.youtube.com/watch?v=3cvP7yJotUM&ab_channel=OlivioSarikas

[vocal burrow]

vietnam flashback

[heady pine]

Easy and to the point

[hollow schooner]

my older one just stoped working halfway so im doing this

[ornate elk]

Hes using AMD so that tutorial won't help

[heady pine]

Ah, my bad then

[hollow schooner]

okay so no error when i open the webui

[ornate elk]

Try to generate on image at default settings.
The first image will take 15-40 minutes
The CMD shouldn't show any errors.

[hollow schooner]

Oki

Just hoping it opens up Firefox

Should i delete my older stable diffusion and move the Lora and models into this one?

[ornate elk]

Yes, if everything works you can do this

[hollow schooner]

Feel like might take up to much space

Oki

[vocal burrow]

I'd wait for the first image to be generated correctly before deleting anything

[hollow schooner]

Yeah ur right

[heady pine]

I meant uninstalling Git and Python as well as the stable-diffusion-webui; it might be a screw around, but it helped me out a lot when I started.

[hollow schooner]

I had to add a bunch of stuff my webui user stuff

Amd problems ig

CS1o helped me a lot with that

It does currently say it launching but it hasn’t opened Firefox yet

[vocal burrow]

screenshot the console

[hollow schooner]

Nvm now it’s downloading something

[vocal burrow]

it's gonna install a lot of things, depending of your configuration / network speed, it can take from 5 to 40 minutes

it has to download a few gbs of data and libraries

[hollow schooner]

Op it launched

Okay we’ll be back in and hour if the image doesn’t generate

Thanks for the helpppp!

[ornate elk]

Alright, no problem 🙂

[hardy rapids]

Hi guys new to automatic1111 after installing it im trying to install the controlnet extension
but i get this error then after rebooting then ui dont even come out anymore. how do i fix this error

[ornate elk]

Hey, did you only installed the Controlnet extension?

[novel basin]

Hello I am having a problem when I try to install the Controlnet extension, already delete and reinstall webui multiple times

[hardy rapids]

yes, getting failed to install handrefinerportable

[novel basin]

looks like we ran into the same issue

[hardy rapids]

yah

tried to google but cant really find the solution

[novel basin]

same here, I kept reinstall, downgrade the pydantic, not helping at all

[ornate elk]

@hardy rapids @novel basin you guys could try to install python 3.10.11 64bit, then delete the venv folder and relaunch the webui-user.bat

[hardy rapids]

3.10.11 is fine with automatic1111?

[hollow schooner]

Am bac it worked thxxxxx

[ornate elk]

Perfect 🙂

[novel basin]

so I just delete the python folder in the ..\webui\system and install the 3.10.11 64 in the same location?

[hardy rapids]

i dont think there is a python folder there unless you are using something else like stability matrix

[novel basin]

interesting, I see, I probably have different UI

[hardy rapids]

yah probably im using this one it didnt have a python folder. from the git pull
https://github.com/AUTOMATIC1111/stable-diffusion-webui

[novel basin]

I download the v1.0.0-pre, am I at the right place?

[ornate elk]

That's not the recommended way to install it

In the pinned messages of this channel you'll find the Guides for Nvidia and AMD cards

[ornate elk]

Hab you tried with 3.10.11 ?
And deleting the venv folder

[hardy rapids]

yah but im getting this error now when i start up.

[vocal burrow]

something might have gone during the install, try reinstalling in a proper folder (not desktop, not any kind of sync'd folder, somewhere with no space/special character in the path, etc)

you can also try to simply delete the venv folder

[hardy rapids]

got it will try tomorrow thanks!

[rare wyvern]

Hello, can i have some advise about making an infinite zoom please. I created a couple but the look junk. I use A1111, would comfyUI do a better job. is there any recommended tutorials to watch to create a good looking infinite zoom, thank you

[hollow jacinth]

Hi community, my buddy tried installing SD and this popped up for him when he tried to launch it. Any thoughts/advice?

[vocal burrow]

no python install detected

make sure they have a compatible python (3.10.6 -> 3.10.11) and that they checked the "add to PATH" checkbox during the install

[hollow jacinth]

yeah, i did ask him if he had it installed, and he mentioned he did install (probably wrong version). I will let him know this.. thanks for the response.

[night elbow]

Hi I'm the friend, (make sure they have a compatible python (3.10.6 -> 3.10.11) and that they checked the "add to PATH" checkbox during the install). I did make sure 'add PATH" was checked. Still not working

Any advice please

Thanks so much in advance

Hi community, I'm trying to installing SD and this popped up when I tried to launch it. Any thoughts/advice?

I've followed all the correct Install instructions so not sure why I should be having trouble

[ornate elk]

Hey, uninstall (not delete) every python version you find.
Then install python 3.10.11 64bit again and check "add python to path"

[obsidian carbon]

ooo boy, do I have a new issue thats a doozy

tl;dr, Stable diffusion is now freezing entire pc when I switch models now ;-; to a point I have to restart pc as it doesnt unfreeze

So, may need to talk in voice to explain as I would be writing an essay here ;-;

[vocal burrow]

how much ram do you have ?

[obsidian carbon]

20GB babes and it shits itself

I cleared the cache which had 5GB x-x

[ornate elk]

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

[obsidian carbon]

Gpu is an Nvidia 1070 ti, inside the command args is this here

[ornate elk]

Okay remove --upcast-sampling and instead add --medvram-sdxl --no-half-vae
Also why are you caching stuff in a folder?

[obsidian carbon]

The cache in the past caused issues on my sdxl but I will try it without

[ornate elk]

Yea try without. You used sdxl without --medvram-sdxl so thats most likely your problem

[obsidian carbon]

Wish me luck 👌

I’m making sure to clear my ram right after also as that seems to help deal with the ram end of things

[vocal burrow]

you have 20gb of ram ?!

how?

16+4 ?

8+8+2+2 ?

[obsidian carbon]

3 sticks ((not optimal i know)) and yes, 2x 8Gb and one 4Gb BUT I have two new 16GBs in transit rn so that will help with ai as well

[vocal burrow]

ok so that's definitely an usual setup but it should be ok.

Did you change any settings to keep more than one model in ram ?

Is your pagefile for c: set correctly ? (aka on automatic or at least 16gb)

[obsidian carbon]

Page file ? Still a noob in the learning face, enlighten me, what is that

[vocal burrow]

when your ram is completely filled, windows start using your disk instead. Moreprecisely it stores everything in the "pagefile".

That's why it needs to be set to something big enough or to automatic (and then make sure you have enough space on c: for it to grow as needed)

Otherwise if there's no pagefile and the ram is saturated your system can become unstable.

[obsidian carbon]

Okay now that , THAT would explain a whole fucking lot ((pardon my swear)) as for the life of me I couldn’t understand why disk was at 100% when I switched models !!

Where is the page file located and I’ll sort that out ??

[vocal burrow]

how to increase pagefile ? https://www.thewindowsclub.com/increase-page-file-size-virtual-memory-windows

Might as well make sure to turn off any unnecessary programs when generating to save up on ram

aka don't run 100 tabs of google chromes, don't game, etc

[obsidian carbon]

Oh that I get as when I use SDXL I literally have the sdxl tab and one yt video and that’s all, everything else is OFF

[obsidian carbon]

I’m curious to how this works as I’ve watched a couple yt videos and a couple of the comments are saying it’s as the same as “Downloading ram” and all that jazz

[vocal burrow]

youtube comments are .... even less reliable than chatgpt

[obsidian carbon]

That’s a fair point to be honest. I’ll give it a go anyway and I can always reset it back to automatic anyway.

What would you suggest I put in for the initial and maximum? Since I’m working with 20GB here

[vocal burrow]

it's a well known mechanism. Linux had a similar mechanism called "swap" partition for ... I don't know ? I'd say 30 years at least ?

[vocal burrow]

16gb should be enough

[obsidian carbon]

For both initial and maximum?

[vocal burrow]

after all that's only """""just in case""""

[obsidian carbon]

It’s reading as MB, what’s the conversion rate to figure that out ?

[vocal burrow]

1024 mb <=> 1gb

EIther you let everything on automatic and make sure you always have at least 20gb free on your c: so it can expand

or you set it to 16gb (min and max) and forget it

[obsidian carbon]

16,384 I think is the mb and I think I’ll set it to 16 otherwise it’ll shit itself lel

[vocal burrow]

The point of this system is to make it more stable. If windows ever need to use it, it will be SLOW, everything will start to lag and maybe even become unresponsive.... But at least it should not crash

[obsidian carbon]

Okay I’ll keep that in mind ^^

[humble cobalt]

hello, tried to use dreambooth but even with settings xformers + fp16 and 8bit AdamW optimizer it says "Exception training model: 'No executable batch size found, reached zero.'."

[rugged summit]

does anyone know why is hires taking so long to upscale at abt 80s/it? using R-ESRGAN 4x+anime6B,15 steps, denoising 0.2, automatic1111 FORGE, 512x512 x 2

[orchid comet]

So.. as just a normal user that wants to try SD3, any easy way I can use this API?

[fair oxide]

ok. been trying to resolve this one my own for a while. can i please get some help resolving these errors when trying to get SD running:

Using cached https://download.pytorch.org/whl/mpmath-1.3.0-py3-none-any.whl (536 kB)
Installing collected packages: mpmath, urllib3, typing-extensions, sympy, pillow, numpy, networkx, MarkupSafe, idna, fsspec, filelock, charset-normalizer, certifi, requests, jinja2, torch, torchvision
Successfully installed MarkupSafe-2.1.5 certifi-2024.2.2 charset-normalizer-3.3.2 filelock-3.13.4 fsspec-2024.3.1 idna-3.7 jinja2-3.1.3 mpmath-1.3.0 networkx-3.3 numpy-1.26.4 pillow-10.3.0 requests-2.31.0 sympy-1.12 torch-2.1.2+cu121 torchvision-0.16.2+cu121 typing-extensions-4.11.0 urllib3-2.2.1
WARNING: There was an error checking the latest version of pip.
Traceback (most recent call last):
File "C:\AI\stable-diffusion-webui-forge-main\launch.py", line 51, in <module>
main()
File "C:\AI\stable-diffusion-webui-forge-main\launch.py", line 39, in main
prepare_environment()
File "C:\AI\stable-diffusion-webui-forge-main\modules\launch_utils.py", line 431, in prepare_environment
raise RuntimeError(
RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check
Press any key to continue . . .

the "runtimerror" at the end seems like maybe the main thing keeping it from running but i also want to know is the 'WARNING: There was an error checking the lastest version of pip" a problem

[ornate elk]

hey whats your gpu?

[fair oxide]

rtx 3070

[ornate elk]

okay, whats your python version? and from where di you installed python?

[fair oxide]

python is 3.10

i downgraded to fix another error

[ornate elk]

3.10.?

[fair oxide]

i researched it and the problem seemed common they were saying python 3.12 doesnt support torch

[ornate elk]

3.10.11 would be great

[fair oxide]

i think i have 3.10.12 actually

[ornate elk]

would work too

[fair oxide]

ok so

RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check

ive tried putting this many places nothing works

[ornate elk]

dont add that, it will use cpu instead

[hardy rapids]

still having the failed to install Handefinerportable after trying to install the controlnet extension in automatic1111
tried a fresh install in C drive with no dashes or symbols in folder and using python 3.10.11

[fair oxide]

i mean its kinda cool these barriers for entry exist. the less people can manage to set this up the more valuable my hack work will be 🙂

[fair oxide]

so what do i do 😦

[ornate elk]

@fair oxide as forge didnt got updated since a month i would recommend you to install the normal Automatic1111 webui.
You find my complete install guide in the pinned messages of this channel. use the guide for nvidia

[fair oxide]

ah thats a good idea

[ornate elk]

if you get that to work, forge could also work

[fair oxide]

yeah

i may switch to a1111 anyway it seems more popular

i had to reinstall windows due to a technical problem i was having

i think i actually talked to you about it last week

stuttering everywhere

[ornate elk]

also make sure you have the latest nvidia driver installed

[fair oxide]

i thought SD may have contributed to it somehow

but we determined my OS had corruption

but i ended up using a vanilla windows 10 usb since the proprietary lenovo boot was giving me shit

so im using windows 10 now

[ornate elk]

ohhh you had the system error that sfc scannow couldnt fix

[fair oxide]

yes

[ornate elk]

a clean install is always better than using inbuild options

discord has massive internet connection errors for me right now

[hardy rapids]

still getting this error when trying to install controlnet on sdwebui
Warning: Failed to install handrefinerportable. Some processors will not work.
Installing sd-webui-controlnet requirement: depth_anything
Installing sd-webui-controlnet requirement: dsine

[fair oxide]

the good part is both problems i was having seem fixed

[ornate elk]

delete the controlnet extension from the extensions folder.
then delete the venv folder.
After that relaunch the webui-user.bat
On the Webui go on the extensions tab, click on Available, click on "Load from"
search for sd-webui-controlnet
and click install. Wait a bit and then close the webui and relaunch

[ornate elk]

thats good ^^, so try the auto1111 install, i think it will work

also auto1111 is the most used webui, forge is still a bit buggy at times

[fair oxide]

does forge even have any advantage at all

[ornate elk]

yes it has some vram optimisations so that it uses less vram and performs a bit better depnding on the gpu.
mostly its a good alternative for nvidia gpus with 6gb or less vram

[fair oxide]

mine has more

8gb

RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check

[ornate elk]

you have the latest nvidia driver?

can you show me the full error in cmd ?

[fair oxide]

thought i did. just resintalled everything

venv "C:\AI\stable-diffusion-webui\venv\Scripts\Python.exe"
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Version: v1.9.0
Commit hash: adadb4e3c7382bf3e4f7519126cd6c70f4f8557b
Traceback (most recent call last):
File "C:\AI\stable-diffusion-webui\launch.py", line 48, in <module>
main()
File "C:\AI\stable-diffusion-webui\launch.py", line 39, in main
prepare_environment()
File "C:\AI\stable-diffusion-webui\modules\launch_utils.py", line 386, in prepare_environment
raise RuntimeError(
RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check
Press any key to continue . . .

and this is what i get after completing step 6 of your walkthrough

[ornate elk]

do you have an intel cpu?

[fair oxide]

amd

[ornate elk]

can you check in the device manager under graphicsadapter if there is 1 or two gpus listed

[fair oxide]

just 1

rtx 3070 laptop

[hardy rapids]

so i tried on stability matrix instead i deleted the controlnet extension folder and venv just to see but getting same error. but i do see some error about permission maybe has something to do with it? how do i fix the permission issue

ERROR: Could not install packages due to an OSError: [WinError 5] Access is denied: 'C:\Users\Tuan Do\Desktop\StabilityMatrix\Data\Packages\stable-diffusion-webui\venv\Lib\site-packages\cv2\cv2.pyd'
Check the permissions.

Warning: Failed to install handrefinerportable. Some processors will not work.
Installing sd-webui-controlnet requirement: depth_anything
Installing sd-webui-controlnet requirement: dsine

[ornate elk]

Okay, open up the nvidia control pannel, on the right click on 3D settings. Then on the left there should be a drop down. There select "Nvidia high performance device"

[ornate elk]

Probably because you installed the webui on desktop

In my guide I say to not do that

[dense tapir]

Hey, would someone redirect me to everything I'd need to install to make videos using XL models in A1111?

[hardy rapids]

i have a regular sd webui installed installed on just C: drive with no wierd names or symbols but same result.

[ornate elk]

Okay no dropdown

That's okay too

[fair oxide]

yeah not seeing exactly as you described

[ornate elk]

Do the following then:
Open up a CMD and type
Pip cache purge
Then delete the venv folder

Then relaunch the webui-user.bat

[ornate elk]

And if you delete the venv folder And the extension in that install?

[fair oxide]

[notice] A new release of pip available: 22.2.1 -> 24.0
[notice] To update, run: C:\AI\stable-diffusion-webui\venv\Scripts\python.exe -m pip install --upgrade pip
Traceback (most recent call last):
File "C:\AI\stable-diffusion-webui\launch.py", line 48, in <module>
main()
File "C:\AI\stable-diffusion-webui\launch.py", line 39, in main
prepare_environment()
File "C:\AI\stable-diffusion-webui\modules\launch_utils.py", line 386, in prepare_environment
raise RuntimeError(
RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check
Press any key to continue . . .

🙂

https://tenor.com/view/snow-white-clap-crazy-happy-insane-gif-17802121

[ornate elk]

whats the full cmd log?

[fair oxide]

venv "C:\AI\stable-diffusion-webui\venv\Scripts\Python.exe"
Python 3.10.6 (tags/v3.10.6:9c7b4bd, Aug 1 2022, 21:53:49) [MSC v.1932 64 bit (AMD64)]
Version: v1.9.0
Commit hash: adadb4e3c7382bf3e4f7519126cd6c70f4f8557b
Traceback (most recent call last):
File "C:\AI\stable-diffusion-webui\launch.py", line 48, in <module>
main()
File "C:\AI\stable-diffusion-webui\launch.py", line 39, in main
prepare_environment()
File "C:\AI\stable-diffusion-webui\modules\launch_utils.py", line 386, in prepare_environment
raise RuntimeError(
RuntimeError: Torch is not able to use GPU; add --skip-torch-cuda-test to COMMANDLINE_ARGS variable to disable this check
Press any key to continue . . .

[ornate elk]

how much free space do you have on C ?

[fair oxide]

a lot

2.94 tb

[ornate elk]

ok

[fair oxide]

we never did anything with the nvidia control panel

[ornate elk]

and the gpu driver install worked without problems?

[hardy rapids]

same

[ornate elk]

thats not like in my guide, you need to create a folder on C, there you git clone the repo in, if not you will get permission errors

[fair oxide]

as far as i know. ill double check it now

[dense tapir]

Hello my dear sir yes hi hello

[ornate elk]

what sort of videos?

[dense tapir]

Adetailer videos, for some reason it kept telling me that I am missing a motion module, even though I placed the xl beta into the same folder which the 1.5 one was in (the 1.5 one does work too), so I'm not sure what the issue might be

[solemn atlas]

where to put LYCORIS

[dense tapir]

Lora folder

[solemn atlas]

thank you

[ornate elk]

Adetailer hasnt a video function, maybe you mean AnimateDiff ?

[dense tapir]

Yes! Thanks for the correction lol, usually mix the names

[ornate elk]

ok, can you show me the error you get?

[fair oxide]

i actually did need a gpu update...

i guess i assumed because i just did a reinstall and i had a gpu driver, that it would obviously be the latest one

silly me

[ornate elk]

oh wow xD

the one that installs itself is not the latest needed

[dense tapir]

Like this seems fine

Adetailer is disabled so Idk what it's yapping about

[ornate elk]

okay, but does it still work?

[dense tapir]

I'm not getting a video, only an image with the prompts I put

Video only works with 1.5 models

[humble cobalt]

Hello, i tried to use textual inversion but when i want to create embedding it says "TypeError: list indices must be integers or slices, not str", i restarted my pc and removed --medvram-sdxl because i saw it worked for someone but doesnt work for me.

[fair oxide]

yeah i guess that was the only problem actually

derp

https://tenor.com/view/brian-regan-derp-what-weird-crazy-gif-11376849

so this is automatic1111?

it looks just like forge lol

[ornate elk]

forge is based of automatic xD

[humble cobalt]

idk if i need to say more informations about it

[ornate elk]

gpu model, and what model you loaded

[humble cobalt]

i have rtx 3060 ti, and for the model i think it doesnt load anything ? like its just creating the embedding to train

but if it matters my checkpoint was set on pony V6XL

idk why it does that because its just "name, initlization text and number of vectors per token"

[ornate elk]

hmm cant help much with training sry :/

[humble cobalt]

okay np

[ornate elk]

i tested it rn and yes, sdxl only outputs an image

[dense tapir]

Hmm okay thanks, is there perhaps another version of Animatediff that we're missing?

[ornate elk]

i got the error that the webui blocks the file because its a .ckpt

trying to convert it to safetensor now

@dense tapir got it working

[dense tapir]

!

[ornate elk]

thats the problem. the webui blocks .ckpt files as security measure.
To fix that:
i copied the motionmodule into the models/stable-diffusion folder.
then i editet the webui-user.bat and added --disable-safe-unpickle
then i went into the Checkpoint merger tab. there i loaded the motionmodule as A.
then selectet "No Interpolation" and on copy config from set it to "None"
then selectet checkpoint format Safetensor. Then press Merge.

After that you copy the new motion module (that is now a .safetensor file) from models/stable-diffusion back to the folder from AnimateDiff
Importan. Close the webui and remove the --disable-safe-unpickle from the webui-user.bat again.
then save and relaunch.

[dense tapir]

Thank you for your help, I'll probably ping later for clarifications, hope that's ok

[hardy rapids]

what does this mean its failing to open webui

[ornate elk]

sure, but later im sleeping ^^

[ornate elk]

do you have any third party antivirus running or vpn or proxy?

[hardy rapids]

nope just the default av that came with windows 11

[ornate elk]

Hmm okay, any browser adblocker?

[hardy rapids]

yah i have adguard

[ornate elk]

Make sure the webui is whitelisted in it

[hardy rapids]

didnt work. i added already. webui opens fine before. but stops working after i try to install the controlnet

after i delete the controlnet in the extesion folder and venf folder it works again. but if i try to reinstall the controlnet i get the issue

[ornate elk]

Maybe you should open up a issue on the Controlnet github page

The dev could know how to fix this

[hardy rapids]

okay got it. thanks

[hardy rapids]

okay i found someone with the same problem in the controlnet github.
https://github.com/Mikubill/sd-webui-controlnet/issues/2757
author said: Please report that in main A1111 repo. ControlNet does not have recent change against how we handle dependency. It seems like pydantic installation failed from your log message due to OSError:

[ornate elk]

Okay, hopefully someone made an issue there then

But good to know its a known issue now

[valid cloud]

Can someone help me with this error? I've tried doing a fresh install and even reinstalling the OS but this keeps showing. My GPU is an AMD RX 6700 XT, 12gb VRAM. Models won't load either.

I'm using Stable Difussion AUTOMATIC111

[ornate elk]

Hey, first of all don't install anything into a OneDrive Folder. That will cause issues.
Secondly: there is a new method for AMD cards using the automatic1111 webui but with much better performance and less vram errors.
Its called the AMD zluda guide. And you find it in the pinned messages of this channel.
For any questions feel free to ask.

[valid cloud]

Thanks, will look into it

[ornate elk]

Its a bit more complex to install but totally worth it.

[valid cloud]

But you got any idea with that error is happening? First time it happens to me

[ornate elk]

I've seen that error before. But your screenshot isnt a full error log of the CMD. Also I would need to know which model and settings you used

[valid cloud]

The one that comes by default, haven't even downloaded any model yet. v1-5-pruned-emaonly

Haven't touch any settings

[ornate elk]

Maybe you didn't edited the webui-user.bat with the correct cmd args

You'll also find the AMD Guide without zluda in the pinned messages.
There are also the needed cmd args listed.

But I would recommend you use the new guide with zluda.

I'm off now gn

[wraith pewter]

WHY?

[wraith pewter]

yes everything went to shit for me as well when i downloaded controlnet

[wraith pewter]

ok im retaredd

[hardy rapids]

i think i found a solution. im trying right now
https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/15551

[wraith pewter]

having a whole bunch of other issues now

how to completely wipe out every SD file from ur pc? cause reinstalling doesnt do shit

[hardy rapids]

you just delete the folder and put in fresh copy then run webui-user.bat

[wraith pewter]

did it work

[hardy rapids]

yes it worked

[wraith pewter]

ok and if u restart it again?

[hardy rapids]

yes it opens and work after restarting

[trail sigil]

Didn't work

[valid cloud]

which one should I add?

[hardy rapids]

how do i save my settings/make template in a1111.
like my tiled diffussion and tiled vae. settings
i tried the png info but it only save my prompts and basic stuff

[deep canyon]

any1 know how to fix?

auto1111, just finished install, 7900XTX (reinstalled cuz new os and no worky :( )

AttributeError: 'NoneType' object has no attribute 'lowvram'

[valid cloud]

same mprobolem

[deep canyon]

@ornate elk silent ping for whenever you're free, since you are the god of Auto1111

[valid cloud]

Did the Zluda guide

[deep canyon]

rip

[valid cloud]

but won't generate anything

What could be the problem?

[deep canyon]

no idea

im using directml on windows

[valid cloud]

I was using directml also

but the lowram error was there

nvm, it worked

[deep canyon]

what did you do

[novel basin]

hello Tuanny
did you follow this solution??

After installing Webui - close it.
Go to A1111 folder
open cmd
type in venv\scripts\activate
When you see (venv) X:\path-to-webui\stable-diffusion-webui>
pip install insightface
pip install --upgrade pydantic==1.10.14
pip install --upgrade albumentations==1.4.2
Then deactivate venv by command
venv\scripts\deactivate
From now run A1111 and you can install controlnet Enjoy!

[hardy rapids]

yes followed that on a fresh install and got it working
when you type cmd and venv\scripts\activate
it may say something about you dont have permission
in that case open windows power shell as ADMIN then enter
set-executionpolicy remotesigned
Press A (Yes to all)
then try cmd venv\scripts\activate and it should work

then just follow
pip install insightface
pip install --upgrade pydantic==1.10.14
pip install --upgrade albumentations==1.4.2
Then deactivate venv by command
venv\scripts\deactivate

[novel basin]

Thank you for the thorough explanation

[ornate elk]

Would recommend to install the directml webui with zluda

You'll find a complete guide in the pinned messages of this channel

[mighty tulip]

For comyui, and I have AMD gpu. Do I need to click the top option?

A youtuber said "if you run it on your cpu it will be way slower than gpu"

However I don't have nvidia gpu, could I still try to run comfui on the "nvidia gpu" file option even if my hardware is AMD?

[hardy rapids]

how do i save my settings in a1111.
like my tiled diffussion and tiled vae. settings
i tried the png info but it only save my prompts and basic stuff

[trail aurora]

Applying attention optimization: Doggettx... done.*** Error completing request*** Arguments: ('task(qc648k7ss7jwouq)', '8x8Pixel', '0.00001', 1, 1, 'D:\\8x8_PixelSD', 'textual_inversion', 512, 512, False, 500, 'disabled', '0.1', False, 0, 'once', False, 100, 100, 'style_filewords.txt', False, 'Cat', '', 20, 'DPM++ 2M', 7, -1, 512, 512) {}Traceback (most recent call last):File "D:\sd.webui\webui\modules\call_queue.py", line 57, in fres = list(func(*args, **kwargs))File "D:\sd.webui\webui\modules\call_queue.py", line 36, in fres = func(*args, **kwargs)File "D:\sd.webui\webui\modules\hypernetworks\ui.py", line 25, in train_hypernetworkhypernetwork, filename = modules.hypernetworks.hypernetwork.train_hypernetwork(*args)File "D:\sd.webui\webui\modules\hypernetworks\hypernetwork.py", line 483, in train_hypernetworkhypernetwork.load(path)File "D:\sd.webui\webui\modules\hypernetworks\hypernetwork.py", line 250, in loadself.layer_structure = state_dict.get('layer_structure', [1, 2, 1])AttributeError: 'NoneType' object has no attribute 'get'

Sorry for bloating chat with that massive error message but I am trying to train my own model using hyperplex and pressing generate nothing happeneds and that error gets posted into the console.

I also seem to be messing the preview prompt which I think may be the cause

[deep canyon]

sadge

[ornate elk]

You need to recheck the path of the zluda folder

In the system variables

[ornate elk]

Hey, no, comfyui doesn't support AMD cards that well. What's your GPU model?

[deep canyon]

I did that before even trying to run it

[ornate elk]

Okay good.
Have you restarted the PC ?
Also latest AMD Driver installed?

[deep canyon]

havent rebooted since my b

shouldve tried that

[ornate elk]

Yea try that. Also do you have an AMD or Intel CPU?

[rustic oasis]

hi everbody ，
Is it because of heat dissipation that Tesla M40 can only use 20% power?

[deep canyon]

5800X3D, 32GB DDR4, 7900XTX

Tuf x570

its now stuck launching lol

different problem means progress as i always say

[ornate elk]

Stuck?

[spiral ermine]

Since you're on the topic of AMD gpus, is there a better solution for using a 5700 XT on Windows than direct ml? It seems that Zluda only works with 6000 and 7000 series cards, and ROCm only in Linux?

[ornate elk]

Hey, zluda works with the 5700xt and is highly recommended instead of directml.
You can follow my install guide in the pinned messages of this channel

[deep canyon]

cant break either

[ornate elk]

Let it load

[spiral ermine]

Any chance it works with Fooocus?

[deep canyon]

also am i supposed to have the --skip-torch-cuda-test? i know normally yes but idk if it can with ZLUDA

i havent used this since directml was the only option

[ornate elk]

No, we need torch and cuda for zluda so that command would prevent it

[deep canyon]

thats prolly the issue then lol

[ornate elk]

Why did you added it?

[deep canyon]

got the error earlier before i rebooted

[ornate elk]

Yea totally

[deep canyon]

still stuck but ill give it a min ig

never took this long when it was just directml

[ornate elk]

You now you need to close everything.
Then open up a cmd and type
Pip cache purge
Then delete the venv folder. And remove the --skip-torch-cuda-test
Then relaunch the webui-user.bat

[ornate elk]

I don't think so

[deep canyon]

wilcom

[spiral ermine]

With this guide, do I already have to have Automatic1111 installed, or is this ok to use starting from scratch?

[ornate elk]

Nope this is from scratch

[spiral ermine]

Ok cool, I'll give it a try in the next few days, thanks

[deep canyon]

also, should I switch the tag from medvram to something else

[ornate elk]

Yea, no problem 🙂 feel free to ask if you get any questions

[deep canyon]

i do have 24GB to leverage

[ornate elk]

You can remove --medvram-sdxl
It was needed before but yesterday I made a few tests with my 7900xtx and it works now good without it

[deep canyon]

baller ty

havent been able to use zluda yet

gonna be interesting

[ornate elk]

Oh you will be amazed

Going from 4-6it/s on directml to 16it/s with zluda is crazy

[deep canyon]

zamn

whats this xformers stuff

do i need that?

[ornate elk]

No more out of vram errors.
You can upscale sdxl by 2 with any amount of Lora's and prompt length.
And you can also use AnimateDiff and it will use 22/24gb xD

[deep canyon]

yeah the out of vram errors were annoying asf

[ornate elk]

Its for nvidia only

[deep canyon]

its trying to launch again but seems as though it might still be stuck

wtf

compooter why

[spiral ermine]

I would be very happy with 4-6 it/s, currently I'm upside down at 58.54s/it

[ornate elk]

Let it load. Its probably dosing something in the background
Can take up to 10-15 minutes

[deep canyon]

Ok

it is using 8-9% of my cpu so yea thats prolly right

yep now its downloading a checkpoint

[ornate elk]

A user who got zluda working with 5700xt tested the speed and got to 1.15 it/s

[deep canyon]

brb bouta ln2 my xtx for more it/s

sadge

fml

[ornate elk]

Can you screenshot the whole cmd?

[deep canyon]

relaunching didnt help

[ornate elk]

Okay, go into the models/stable-diffusion folder.
There delete the 1.5 EMA pruned model.
Then download the dreamshaper v8 from here and place it in the same folder.
https://civitai.com/models/4384/dreamshaper

Then relaunch

[deep canyon]

trying now

didnt change anything

shits fucked, yo

it dont wanna work

very big L

just doesnt do anything at all when i input a prompt

not giving the vram thing now just not doing jack

[ornate elk]

That's intended

Read my guide

The first gen will take 15-40 minutes

[deep canyon]

oh right my bad

i honestly forgot about that part

i read it I swear 😭

[spiral ermine]

Oh, I forgot to ask... this is for SDXL, or SD 1.5?

[deep canyon]

its just late and im eepy 😭

[ornate elk]

Np, your 99% there

Now just patience xD

[deep canyon]

https://tenor.com/view/cat-kitty-blep-blehhb-silly-gif-4711505992572152292

[ornate elk]

For 1.5

[spiral ermine]

Dangit

[ornate elk]

But you also can use sdxl faster

[spiral ermine]

How so?

[ornate elk]

I think on directml with 8gb it wouldn't even loaded

[deep canyon]

whats sdxl times looking like on an xtx?

[ornate elk]

4-5 it/s

[spiral ermine]

Fooocus works with my 5700 XT, but I think it's the only SDXL variant that does

[deep canyon]

so i dont actually know what that metric means

just that more is more good

[ornate elk]

Okay, yea I mean sdxl will work much faster on zluda than on any directml based webui

[deep canyon]

4 or 5 iterations per second?

and is each sampling step an iteration or something entirely different

[ornate elk]

Sampling step should be an iteration but I'm not 100% sure

[deep canyon]

okie

[spiral ermine]

But if your guide is for 1.5 and not SDXL, then I don't think there is a way to make Zluda work with SDXL on a 5700 XT

[ornate elk]

My guide is for a webui that can use any model. 1.5, 2.1, sdxl, (sd3 in future)
Sdxl will work. And we can tweak it to use less vram

[deep canyon]

looks like it be doing some work

[spiral ermine]

Oh, so I can just drop in SDXL models and use them instead of 1.5 models?

Ok, I have to update my drivers it seems, but I will work on it and see what happens

[hardy rapids]

upscaled this shiba image in automatic1111. is there a way to color match it to the original image on the left im new to a1111. the upscaled image on the right is a bit darker/higher contrast

[dense sky]

hi huys, i spent allomost the last 2 weeks trying to get kohya_ss up and running
i tried windows 10 as well as ubuntu22, however i cant seem to get it fully working to use the GPUs
anyone has propper instruction especially on driver versions,os versions and cuda versions
i basically tried cuda11.8 which apparenly is allmost uninstallable on ubuntu22 by now because the drivers 520 wont build
with 535 i cant get any real speed however the gpus is utilized
what OS Ubuntu version do you recommend?
and what cuda and driver version?

[ripe bronze]

How do I know which model folder to put these in? Was it on civiait page where I got them from?

[ornate elk]

Everything that is 2gb or bigger goes into the models/checkpoints folder
The smaller files go into models/oras

What webui do you use?

[ripe bronze]

comfyui

Is this what you mean?

the top file is 5gb so it should be in this "models" folder?

[vocal burrow]

models\checkpoints

because it's a stable diffusion checkpoint

(most likely as said in its description page)

[ripe bronze]

2gb is a checkpoint?

[vocal burrow]

probably a lora (but they could be embeddings or vae or something else)

what you're downloading is always mentioned on the webpage somewhere

[ripe bronze]

I see thanks I'll check those description pages

[vocal burrow]

eg

[normal silo]

[novel basin]

thank you all solved

[wraith pewter]

I have 7900xtx too. Do u have full guide for zluda? I keep getting vram errors too

[ornate elk]

Yes in the pinned messages of this channel, you find my guide

[ornate elk]

Let me know if it works or if you have any questions about it

[viral walrus]

Hey, I just downloaded stable diffusion and am using the Comfy UI. I'm following a tutorial with how to use it. However, when I get to the generate image step, I keep getting errors.

I downloaded two different models, the orginal stable diffusion, and a custom model. I get the same error for both models.

The error I receive is 'unknown error.'

What did I do wrong? Can Anyone help?

[elder turtle]

Does the console spit out anything more than unknown error?

Did you download and select that checkpoint?

[viral walrus]

Give me one second

[viral walrus]

I think I solved half the problem. So I tried restarting everything again, but actually closing my browser fully and at least now the default model works. I guess despite me closing the window, and closing the command prompt, it never refreshed or something?

[wraith pewter]

do u need the old a1111 or can i delete that one?

[ornate elk]

you can delete it, but save your models and loras

[wraith pewter]

i also saw theres one in sys32. can u delete that too?

[ornate elk]

Yes please

It happens when you don't git clone into a specific folder

[wraith pewter]

oh lol

[ornate elk]

My guide covers that step so that it won't happen

[pallid pollen]

anyone an idea what causes this inside auto111: RuntimeError: Sizes of tensors must match except in dimension 0. Expected size 1152 but got size 992 for tensor number 4 in the list.

[wraith pewter]

and i need the zluda windows amd zip right?

[ornate elk]

Yes

[ornate elk]

What settings did you used?

[vocal burrow]

shot in the dark you're using a CN 1.5 model with a SDXL model

but without a proper full log and / or settings used we can't know for sure

[wraith pewter]

can i just run one time without a model so it does it thing

[ornate elk]

It needs at least one model

[wraith pewter]

this normal?

[elder turtle]

The last workflow keeps being loaded in comfy

[fair oxide]

Hello!

I nee help

[ornate elk]

Are you one the last step of the guide? First image will take 15-40 minutes

Is the generate button grayed out?

[fair oxide]

I'm having issues with my project
https://vega.js.org/paint

[wraith pewter]

now i get this shit

after restarting pc

[ornate elk]

Can you show me your path settings?

[ornate elk]

Is your AMD driver updated?

[wraith pewter]

i think so? how to check

[ornate elk]

Open the AMD Adrenalin software

Then on the gear icon

[wraith pewter]

it all worked fine when i first installed it

but then i generated without model and it got messed up and then i closed it and it showed me all this errors

[ornate elk]

Do you have a model inside the models/stable-diffusion folder?

[wraith pewter]

yes

the one u recommened

[ornate elk]

Dreamshaper v8 I guess

[wraith pewter]

yup

also did the pip purge and deleted venv

[ornate elk]

When?

[wraith pewter]

5 min ago

[ornate elk]

Before or after the skip torch cuda error?

[wraith pewter]

after

[ornate elk]

Hmm okay

Restart the PC. Then relaunch the webui-user.bat

[wraith pewter]

ok

ok its doing smth new

back up

0%| | 0/20 [00:00<?, ?it/s]

is it supposed to say this tho

when generate

[ornate elk]

Yes it will take a while

[wraith pewter]

ok thnx

ók its working

[ornate elk]

Good 🙂

Have fun with the insane speed ^^

[wraith pewter]

yup its insanely fast

also upscaling

[ornate elk]

Yes hires fix takes seconds

[wraith pewter]

also whats the proper way of closing the program

just quitting the cmd?

[ornate elk]

Yep there is no exit or shutdown

[wraith pewter]

Thanks so much!

[ornate elk]

No problem 🙂 have fun. Also for sdxl get the fp16 sdxl VAE.
And the Tiled VAE extension if you upscale at higher resolution.

[fair oxide]

Someone tell me what to fix

[unique bluff]

Hello again. I’ve tried to install sd again multiple times (I have an Intel Mac, my GPU is Intel Iris Pro Graphics 62000 1536 MB and I have 8GB of ram), but I still get the same ```File "/Users/sweetcake/stable-diffusion-webui/modules/sd_models.py", line 662, in send_model_to_cpu

if m.lowvram:

AttributeError: 'NoneType' object has no attribute 'lowvram' error on the code and'NoneType' object has no attribute 'lowvram'``` when trying to generate an image. Is there any way to fix this? I couldn’t find any answers online

[ornate elk]

Yes, there is a fix.
Delete the 1.5 ema pruned model from the models/Stable-diffusion folder.
Then download the Dreamshaper v8 model from here and put it in the same folder.
Then relaunch the webui.
https://civitai.com/models/4384/dreamshaper

[unique bluff]

Hummm, let me check

[ornate elk]

You also need to select the new model in the dropdown in the webui

[unique bluff]

On that note, I’m not able to do that either

(At least with the 1.5 pruned and other models on the folder)

[ornate elk]

Are the other models at least 2gb in size?

[unique bluff]

I had one 3.85 GB

That didn’t work when I just had it nor with 1.5 pruned

Whenever I tried to switch the model it would make the loading part and then proceeded to not change the model

Sometimes it won’t even put the only model available on the folder

[ornate elk]

Okay that's strange

[unique bluff]

[ornate elk]

Do you have a 2gb model?

If not try the one from the link

[unique bluff]

I could try to record what happens, it’s a bit complex to explain properly

[unique bluff]

Nop, I’ll try the one from the link

I should also mention I’ve been trying to use sd without a VAE

Not sure if it interferes

[wraith pewter]

@ornate elkstarted doing this on its own all of a sudden?

oh shit it just exactly described the image

thats awesome

[ornate elk]

Thats ok, it should still work

[unique bluff]

It didn’t

🤨

[ornate elk]

Have you used the install script for mac from the pinned guides?

[unique bluff]

Yes

In the last time I installed it I used that script (in the times before I didn’t)

[ornate elk]

Hmm okay

[unique bluff]

I was able to find the lines where the error is happening and it’s def send_model_to_cpu(m): if m.lowvram: lowvram.send_everything_to_cpu() else: m.to(devices.cpu)

[ornate elk]

This lowvram error pops up quite a lot recently

What's in your webui-user.sh?

[unique bluff]

The code?

#########################################################
# Uncomment and change the variables below to your need:#
#########################################################

# 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"

###########################################```

[ornate elk]

Hm there are no variables set

No launch args

But maybe the script starts the webui.sh with some additional args

[unique bluff]

Humm, should there be launch args?

I also have a webui.sh and a webui-macos-env.sh

[ornate elk]

Need to be Afk. I will check the script later

[unique bluff]

Ok

[keen edge]

I need help with this, anyone?

[ornate elk]

You are using Forge webui.
It has Controlnet already integrated.
You see it in the txt2img tab when you scroll down

[keen edge]

Hm, makes sense but, why I can't select any models for openpose? I installed several of them

[ornate elk]

yea its pretty buggy in forge

most people link their controlnet models of their auto1111 installation

[fair oxide]

Hello. Help me please. When I try to use the Train hypernetwork function, I get this error. What to do?

I didn’t change anything in the files, I just installed CivitAI Browser, lycoris and CivitaiHelper. I found this on GITHub, but the link inside is dead: https://github.com/AUTOMATIC1111/stable-diffusion-webui/issues/13080

Error:

[crimson sail]

Can anyone help me out with this?

Error occurred when executing SaveImage:

Numpy is not available

File "O:\stable-diffusion-comfyui\ComfyUI\execution.py", line 151, in recursive_execute
output_data, output_ui = get_output_data(obj, input_data_all)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "O:\stable-diffusion-comfyui\ComfyUI\execution.py", line 81, in get_output_data
return_values = map_node_over_list(obj, input_data_all, obj.FUNCTION, allow_interrupt=True)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "O:\stable-diffusion-comfyui\ComfyUI\execution.py", line 74, in map_node_over_list
results.append(getattr(obj, func)(*slice_dict(input_data_all, i)))
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "O:\stable-diffusion-comfyui\ComfyUI\nodes.py", line 1407, in save_images
i = 255. image.cpu().numpy()
^^^^^^^^^^^^^^^^^^^

same thing posted in #🧣｜comfy-ui

[viral walrus]

I'm attempting to use Lora Training in ComfyUI. Everything works besides when I get to the Training part, and I keep getting this error. I have tried reinstalling python about three times. Remove it from the microsoft store, found earlier versions on my system and removed it, disabled the app launchers from the app list. I've also attemtped to manually download the missing files as well. Maybe I'm just missing something basic? Any help would be great!

[fair oxide]

@ornate elk sir could you help me pls

[ornate elk]

cant help with API stuff, sry

[fair oxide]

Then whoo 😓

[brittle lantern]

Guys I have a question how do I get access to SD 3.0?

[vocal burrow]

it's only accessible through stability.ai's API so far

[brittle lantern]

Alr

Can u link it and is there a tutorial?

[vocal burrow]

https://platform.stability.ai/docs/api-reference#tag/Generate/paths/~1v2beta~1stable-image~1generate~1sd3/post