Duskfall Crew's picture

Duskfall Crew

Duskfallcrew

AI & ML interests

Dissociative Identity Disorder in AI media, Artist work in AI models, Skipping out on math class, being the brat in the back. Making space for others like me to feel safe to create art in open spaces. Repos are mostly at Earth & Dusk from now on.

Recent Activity

Organizations

Earth & Dusk's profile picture Earth & Dusk SD Diffusers's profile picture Isekai's Screencaps's profile picture Mickey's Diner's profile picture Social Post Explorers's profile picture Hugging Face Discord Community's profile picture

Duskfallcrew's activity

replied to their post 36 minutes ago
view reply

Oh you're welcome to check out the code, and suggest improvements.
This is my running tool i've been using, i started doing an a111 via claude, but the thing is i don't program - I know a TINY BIT AROUND LOGIC because of web design and like early learning of programming before python existed.. .but i'm a designer jim, (star trek reference) not a programmer.

I used google to fix it up, the new gemini one, it's ... actually QUITE easier than gpt or claude for programming, and it's got a FAIRLY GOOD wide "freebie" range, like you get 1million lines of code or text with it, and it's REALLY good.

one of those things that, Llama isn't BAD but it's more for like general use, and i know that sadly gemini isn't really open source, but right now, someone like me -- needs something THAT BEEFY :D

The colab one SHOULD WORK because i actually used it, uploaeded an AI song of mine to one of my personal repositories, i made that tool go on personal rather than E&D -- just cause it's on personal rather than the E&D organization on github.

posted an update about 5 hours ago
view post
Post
142
Just finally ... added my HF backup tool to an HF repo... after two years roughly of making this -
It's inspired KINDA by Camenduru but his one stopped working and i wish i had the original code for it so i could reformat the A111 extension he had...

Becausei 'm TRYING to make an A111 extension and maybe a custom comfyUI node:

Duskfallcrew/Huggingface_Backup

I originally patched this from Everydream2trainer and some other stuff.
So the credits stay.

I'm amazed at this lol.
  • 2 replies
ยท
replied to their post 13 days ago
view reply

"...YNTEC was largely doing backups of diffusers, which is valid in itself, but sometimes that lead to clashes with model creators on Civitai..."

I agree for the part on the CIVITAI.COM. But Yntec also really care about author license, as we discussed on that subject.

I disagree for the part on was largely doing backups of diffusers. Because Yntec always fine-tuned them with a default resolution at 768px instead of 512px, and either VAE baked-in/either LoRAs fusions, I need to say Yntec models are more creative than the original are. Yntec is a passionate person, I mean that he care attention to precise details on each of his models (either conversions from LoRAs either conversions from Diffusers), just try his models, then try originals. With DEFAULT settings not in WebUI, and you will see a great improvements over each original models.
Try testing LadyNostalgia or even Gacha from Yntec, you will certainly be more than pleased to the output result!

Edit : typo fix on my orthograph xD

Yea the hard part is a lot of people didn't agree, i don't mind - i think i'd found some of mine backed up, and converting to diffusers meant other sites had access that a LOT OF US DIDNT KNOW about - but i do know diffuesers saves people time and money and i'm ok with that.

Now if anyone catches CalypsoCrunchies ..... i know nothing

replied to julien-c's post 14 days ago
view reply

Yea i'll DEF consider if my budget can handle it do Enterprise for at least E&D - right now i'm juggling several small income things (none of which can really be classified as a business) and being that i don't really have a job it'll be depending on how i juggle stuff.

replied to their post 14 days ago
view reply

@Yntec will perhaps never come back again... given the way he put all his models offline... ๐Ÿ˜ต. And it's a huge lost of precious models !

By the way, I'm already at 25GB used on my 500 GB of free storage... Just with 8 models lol. ๐Ÿ˜‚... 500 GB is really not a sustainable value for hardware capacity when it comes to store on HF these GB models!

I think, 1TB for free storage and unlimited for PRO would be great!
(if one day, it come to unlimited storage for PRO, I sign immediately for a PRO account each month ๐Ÿ˜‚).

If not... What is the point then.

But.. @Yntec took down his 449 models.. He was really upset. Hey come back lovely @Yntec ๐Ÿ˜†.

YNTEC was largely doing backups of diffusers, which is valid in itself, but sometimes that lead to clashes with model creators on Civitai...

Public storage- y'all ... HF are you nuts? LOL - Thank you for the effort, starting to wonder if i should be kind and at least pay for enterprise if it ain't expensive on E&D -- I dont know if you can classify half the stuff i do as research -
But the idea that public storage for organizations at least is good.

reacted to nroggendorff's post with ๐Ÿš€ 20 days ago
view post
Post
1949
uhh..
ยท
replied to their post 20 days ago
view reply

@reach-vb Hey, totally valid that the staff here DO comment and come back at us with updates. Somedays i feel like making models and throwing them at the wall like playdough isn't really "RESEARCH" but i guess the idea of creation is sort of the other half of research? It's not that i DONT do that stuff, it's that honestly - i have a github full of random python toys i'm making and none of them end up here lol

posted an update 21 days ago
view post
Post
1313
So this thing about quotas?
I checked.
SOMEHOW i'm uhh
Not even close on main, probably over on my org but what can you do?
You never get unlimited.

Never have and never will :D
ยท
replied to nyuuzyou's post 21 days ago
view reply

LOL, 'im not terribly worried, i'm acutally a paid creator on civitai, i just can't gen locally so i HAVE to upload my models up here.

replied to nyuuzyou's post 21 days ago
view reply

The thing is datasets are also semi a copyright issue, they're not just looking at the $$$ but the stupidity of the lawyers who will look at a tshirt and call it porn. (I say this because amazon's rekognition tagger loves to tell me a clothed human is explicit nudity)