【企画】6月4日リグルリレー2023 第2走者 | Encode https://t.co/l38AQQrkp3

29 77

My entry for the The Hidden Realm Week Challenge - today's theme being "Guardians Of The Realm".

3 28

🌕 As the moon casts its gentle glow, let go of your worries and embrace the calm of the night. Sweet dreams, everyone!

0 5

🎨 "Art is not what you see, but what you make others see." - Edgar Degas. Inspire and captivate others with your creations.

1 6

🌱 "The earth without 'art' is just 'eh'." - Anonymous. Let art bring color and excitement to your world.

0 4

🌟 Good morning! Remember, you are capable of achieving incredible things today. Seize the day! 💪🌞

0 2

🌃 Night has fallen, and it's time for our minds to wander in the world of dreams. Sleep well, and see you tomorrow!

0 4

【歌ってみた】ブリキノダンス/su-i

■Vocal&Rec:su-i
■Mix&Encode:Tsuji

YouTube⏬
https://t.co/bKca169lgZ

4月だ!🌸🌸
気持ち新たに早口な歌で勢いつけて
突っ走っていこうね(?)ということで
聴いてもらえたら喜びます‼️
感想をくれたらもっと喜びます‼️


13 31

🌟 As the stars light up the sky, may your dreams be filled with inspiration and peace. Sleep well, everyone!

0 3

TBC trio :3c

May turn these into stickers would anyone be interested?

7 13

ok, here’s some kinda recent wc designs i made!!
(if you count 2021 as recent!🤣)
the TBC protagonists!💖
some details in the 🧵

3 4

🌟 Daily Art Showcase: Check out this mesmerizing piece from today's creation!

2 12

🌅 Rise and shine, beautiful people! A new day awaits, full of opportunities and adventures! ☕️🌞

0 2

🌅 Good morning, art enthusiasts! We're excited to share another AI-generated masterpiece with you today!

1 7

🎨🤖 Introducing the ultimate NFT and A.I. Artist! Follow my journey as I create innovative digital art!

1 3

dynamic-thresholding latents in pixel space.
at sigmas≥1.1: we decode to pixel space, do Imagen-style thresholding, encode to latents.
trained a tiny latent decoder + RGB encoder on VAE outputs (you could call this distillation).
left = CFG 30, usual
right = dynthreshed

5 31

イキテオリマス。auto encoderの自動調整ですが、ここまでいけました。HDRも内部的には対応できてるようなので、exrかhdrで出力できんもんかと模索しております!

0 0

[wd-1-5-beta2](+LoRA)
学習元44枚。キャプションはDD。バッチサイズ=2。repaeats=10。unet_lr=1e-6。text_encoder_lr=5e-8。20epoch。

まぁこんなもんかな。学習足りない感じあるけど、過学習するよりはまし?学習率難しいな。

0 0

diffusersって2系のtext encoderは最終層が既に切除されてて、自動的にclip skip=2になっていると考えていいんだよね?
このissueは読んだ
https://t.co/YhX7f0fjBG

というか、wd15はほとんどclip skipの違いが出ないのか?
1: clip skip = 1
2: clip skip = 10

2 15