don't forget the config file (yaml) file.
Clip skip 1 is recommended, but 2 is also suitable (the style is very different)
I suggest using nick-x-hacker/bad-artist embedding for best results:
https://huggingface.co/nick-x-hacker/bad-artist
as well as prompts like:
(masterpiece, photo, highres, realistic, best quality, hd, high detail, a lot of small details:0.9), (Lopunny, furry), full body, perfect body, large breasts,
Negative prompt: human, simple background, art by bad-artist, lowres, text, error, cropped, worst quality, low quality, normal quality, jpeg artifacts, signature, watermark, username, blurry, bad anatomy, bad hands, ugly, morbid, extra fingers, missing fingers, extra digit, fewer digits, poorly drawn hands, mutation, extra limbs, gross proportions, missing arms, mutated hands, long neck, duplicate, mutilated, mutilated hands, poorly drawn face, deformed, bad anatomy, cloned face, malformed limbs, missing legs, too many fingers,
Steps: 20, Sampler: Euler a, CFG scale: 10, Size: 512x512, Model hash: 3edd27788a, Model: AnythingFurry, Denoising strength: 0.5, Hires upscale: 1.5, Hires upscaler: Latent
send your results or errors in the comments.
Description
FAQ
Comments (27)
Any chance of a more detailed description of what it's a merge of?
The YiffyMix model which is based on yiffy-e18 and Zephyr F111, anything and a few others, but these 2 are the main ones
@XXIKEK Cool
The file was misnamed - as .safetensors instead of .ckpt
testing now.
I cannot load the file, weui reports the file as corrupted :(
"-----> !!!! The file is most likely corrupted !!!! <-----
You can skip this check with --disable-safe-unpickle commandline argument, but that is not going to help you.
changing setting sd_model_checkpoint to anythingfurry_1.ckpt: AttributeError
Traceback (most recent call last):
File "C:\Users\(username)\stable-diffusion-webui\modules\shared.py", line 534, in set
self.data_labels[key].onchange()
File "C:\Users\(username)\stable-diffusion-webui\modules\call_queue.py", line 15, in f
res = func(*args, **kwargs)
File "C:\Users\(username)\stable-diffusion-webui\webui.py", line 84, in <lambda>
shared.opts.onchange("sd_model_checkpoint", wrap_queued_call(lambda: modules.sd_models.reload_model_weights()))
File "C:\Users\(username)\stable-diffusion-webui\modules\sd_models.py", line 428, in reload_model_weights
load_model(checkpoint_info)
File "C:\Users\(username)\stable-diffusion-webui\modules\sd_models.py", line 385, in load_model
load_model_weights(sd_model, checkpoint_info)
File "C:\Users\(username)\stable-diffusion-webui\modules\sd_models.py", line 240, in load_model_weights
sd = read_state_dict(checkpoint_info.filename)
File "C:\Users\(username)\stable-diffusion-webui\modules\sd_models.py", line 223, in read_state_dict
sd = get_state_dict_from_checkpoint(pl_sd)
File "C:\Users\(username)\stable-diffusion-webui\modules\sd_models.py", line 194, in get_state_dict_from_checkpoint
pl_sd = pl_sd.pop("state_dict", pl_sd)
AttributeError: 'NoneType' object has no attribute 'pop'"
@FoxDude strange, I have everything working with .safetensors
anyway send pictures in the comments.
@FoxDude try using .safetensors with config file (yaml)
@XXIKEK I did use the config file, but..
Forgive my stupidity, I never noticed "safetensors" was a stable diffusion filetype, so I renamed it, thinking you'd made a mistake. I tried again without renaming it, and it works great.
Results look excellent so far! Going to bed now though, might check back at a later time.
Cheers
How well can it generate characters with scales rather than wool?
try adding wool to negative prompt, or specify full body (scales) like this for prompt, might work. I tested it.
(scales), highly detailed scales, shiny scales.....something like that works well as I can recall on the spot
Keep getting|
modules.devices.NansException: A tensor with all NaNs was produced in VAE. This could be because there's not enough precision to represent the picture. Try adding --no-half-vae commandline argument to fix this. Use --disable-nan-check commandline argument to disable this check.|
I put the files in the model folder:
stable-diffusion-webui\models\Stable-diffusion
and the config in the config folder:
stable-diffusion-webui\configs
I'm not super competent with programing, when it say I can enter the lines to disable the check where exacly do I put them in or is the something I didn't do right. Thx for help in advance.
The config goes in the same folder as the model (important) - that is - put them both in stable-diffusion-webui\models\Stable-diffusion.
I got the same error as you at one point, I think it was when I was using the "Hi.res" fix/integrated upscaler.
What it suggests you should do, is to run Stable diffusion with that argument.
Put it in your webui-user.bat script - the file you use to launch Stable diffusion.
Contents of the file should be like so;
@echo off
set PYTHON=
set GIT=
set VENV_DIR=
set COMMANDLINE_ARGS=--no-half-vae
git pull
call webui.bat --xformers
I have xformers in there as well, since it's generally recommended for memory optimization.
You could just create another batch script for testing, or for keeping things neat.
Edit the file, and save it as another "*.bat" file.Remember to manually add the ".bat" if you're using notepad, as it may remove it, and simply save it as a *.txt
@FoxDude Thx for the help
Looks like there might be something wrong with the model's VAE. I and others keep getting the error "A tensor with all NaNs was produced in VAE" and using the command line argument "--no-half-vae" does work but the image result is significantly "worse" in my opinion. Using the "--disable-nan-check" argument doesn't really fix anything as it just returns a blank image. Made sure to even start without any command line arguments, but same result with error. Lowering the steps does prevent it sometimes but well it's frustrating. Great model when it works. I do also note that CivitAI does have warning shield on the config file so maybe that's related? I looked through it but didn't catch anything obvious (also I'm pretty new to this).
--was corrected in reply as no VAE so removed the top line
++ new edit, seems to happen mostly when there is a large close to 150ish character prompt.
The model does not have its own VAE, I used another model's VAE and used "Bake in VAE", make sure you downloaded the configuration file and put it in the same folder as the model, right next to it.
Also try using the following startup parameters
set COMMANDLINE_ARGS= --xformers --autolaunch
and then add the next line:
git pull
@XXIKEK I tried a bunch of things like '--xformers, --lowram, --medvram, --disable-opt-split-attention' etc but it seems to mostly have issues at higher steps etc. The config file is there and I made sure to redownload it and check it. It seems to also happen mostly with the Euler samplers so I've just not been using them. Though I do get it usually when I try to "Hires" them. It's been great since though and haven't had another of the same error. Even ran it all night generating 100s of images at 150 steps just to test, so not sure. On a side note, thanks for this model. <chef's kiss> Great work!
Do e621 nsfw tags generate sex scenes for this model?
This is pretty fucking awesome
Any recommended VAE? Since the color is a little bit uncolorful.
For some reason, anything I try to generate with this winds up as random colorful noise. Other checkpoints work fine. Any ideas as to what the issue could be?
EDIT: so, deleting the YAML file worked to fix this, and I'm getting good results, but occasionally, seemingly at random, it will produce fully black images. What's up with that? Anyone else getting it?
FIXED: So, it looks like it was probably a vram issue; I found that increasing the number of sampling steps let the image generate normally.
where do i put the config file and the .part file? using webui. wont show up as an option when generating
You need to download SafeTensor and yaml files and put them in the folder: models -> Stable Diffusion
@XXIKEK i use easy diffusion. were would i put the config file, or is it the same C:\EasyDiffusion\models\stable-diffusion?
@orneliar108193 I use AUTOMATIC1111, but probably yes, you need to perform the same steps
This is so underrated.. is there a list of built in characters that work with this?
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.