琥珀青葉@LyCORIS
@KBlueleaf
Followers
2,003
Following
777
Media
424
Statuses
1,793
Undergraduate in Taiwan. Leader of LyCORIS
Taiwan
Joined May 2021
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Colorado
• 150998 Tweets
Notre Dame
• 126678 Tweets
Edmundo
• 108851 Tweets
#AEWAllOut
• 97929 Tweets
Nebraska
• 50114 Tweets
イコラブ
• 42830 Tweets
T-岡田
• 32820 Tweets
Talleres
• 30955 Tweets
#KPOPMASTERZinBANGKOK2024
• 29989 Tweets
Oregon
• 25904 Tweets
#InternationalLiteracyDay
• 22670 Tweets
Bama
• 18975 Tweets
Willow
• 18139 Tweets
Shedeur
• 17812 Tweets
Swerve
• 17780 Tweets
ショウマ
• 17536 Tweets
セントウルS
• 16939 Tweets
NC State
• 15106 Tweets
DEAR MYLOVE LINGORM
• 15015 Tweets
Danielson
• 12686 Tweets
#LingOrmFanMeet2geTHer
• 12076 Tweets
Deion
• 12007 Tweets
まーさん
• 11861 Tweets
設営完了
• 11400 Tweets
Ospreay
• 10765 Tweets
Pinned Tweet
The paper of LyCORIS project is accepted by ICLR 2024!!!!!!!!
"Navigating Text-To-Image Customization:From LyCORIS Fine-Tuning to Model Evaluation"
My first ever paper, as an undergraduate student.
Arxiv here:
5
20
118
The one of best SDXL anime base model: Kohaku-XL delta is released!!!
It is trained on 3.6M images with LoKr algorithm. And it can generate various style and characters
8
79
306
Just started a part-time gig as an ML Engineer at Apple, collaborating with their research team. Looking forward to working on some cool projects!
10
1
139
@DaDa8964
作為一個"dragon迷" loong這個音譯我倒是覺得可以接受
(dragon, drake, amphithere, wyvern, wyrm.... 分類太多了，然後東方龍/中國龍又跟西方奇幻作品的龍的體系毫不相關)
但如來佛這個太好笑了
1
2
113
The extension of DanTagGen in sd-webui is now done!!
You can try to use it now:
Just write your prompt as usual and it will "upsample" your prompts.
3
30
114
I captioned 7.45M danbooru images with Florence-2 (from msft)
currently only "<MORE_DETAILED_CAPTION>" has been done and "<DETAILED_CAPTION>" is running.
"<OD>" and other targets are planned but low priority.
3
22
112
So the cool thing is here
Basically a special use case of PaintsUndo
I will use this to make sketch-lineart-image pair for reproducing some kind of "copainter"
1
24
98
1
7
79
Kohaku XL Zeta released
better character fidelity
support NL prompt
trained on 8.46M images
consumer level hardware
0
22
83
LyCORIS 2.2.0!
QLyCORIS and DoRA supported.
(bnb 4bit/8bit linear layer with LoHa/LoKr/OFT/BOFT)
(Weight decompose method for LoRA/LoHa/LoKr)
0
15
81
Kohaku-XL alpha published!
Trained on 100kπ imgs and merged with 2 models.
available on civitai now (link below)
2
20
76
DanTagGen-beta is released!!
Pretrained on 5.3M entries with 10epoch.
Same 400m param arch
Better performance!!!
Check the Hugginface repo or HF space:
Thx
@Gradio
for zero-GPU resources UwU
1
17
66
@oAcDyiuHAw4H5eU
@tindingtin
@whyyoutouzhele
成年人被認為應當有一定的自保能力，但未成年人則需由其他監護人協助保護。
未成年人傷亡事件意味著一定有監護人疏失在內。
（這裡的監護人並非單指法定監護人，還同時包括了所在場所的負責人等）
或者說，未成年人傷亡事件暴露出的隱患更多。
0
3
61
0
4
56
My outline expansion method can help existing pixelization algorithm to get way better result.
Check my repo for more informations:
0
11
51
LyCORIS 3.0.0 is finally released
over 200 commits
brand new module/functional/parametrize API, more efficient implementations, better testing and brand new banner
0
11
52
1
0
47
The most unsafe part of AI is not the AI themselves, but their users, and "bad Anit-AI tools"
The orig img is from "human artist".
With some "not generative model" process.
I can fake it as "AI generated image" easily.
Will this tool protect "human artists"? or attack them?
5
7
46
Paint UNDO is good for making synth dataset (for intermediate result of an image)
0
5
44
A full demo of DanTagGen + Kohaku XL is now available on HuggingFace Space!
"This Cute Dragon Girl Doesn't Exist"
1
7
38
Contrast-based downscale is now implemented into Python UwU.
Better color choice when downscale (for pixelart-like style).
Also I add some example of using my outline expansion method to improve other existed Pixelization Method. (Even NN-based method can get improvement)
2
7
35
Just published a metadata database for Danbooru. (Use the official dumps of posts metadata, not by crawlers)
You can now use this database for tagging/analysis or whatever you want.
Model card also provide a link to the dataset which have all the img.
1
6
33
190k step, bs32, 9hour on single 4090.
(Totally from scratch)
actually not bad
3
2
32
You also don't need to spell "aua" to get Houshou Marine in
#KohakuXL_Delta
0
5
32
LyCORIS is now updated to 2.0.0 for lot more new features.
Please check it if you are using lycoris!
0
6
31
@AItyoitashi
たんまさん、21000フォロワーおめでとうございます！
And this is My first time for seminar!
Hope you guys love my works!
#新人AI術師さんと繋がりたい
1
3
27
<DETAILED_CAPTION> for Danbooru2023 is done:
Will use phi3-hornyvision to get the caption in near future.
0
2
27
Just found I already spend 5000 3090hour on kohaku xl and DTG project
With my own GPU only.
(All the broken run and dataset processing are not counted, only training time for published model)
I think this is quite impressive
2
0
27
SimpleTuner now support LyCORIS training!
0
5
27
LyCORIS package is now updated to 2.1.0 UwU
With BOFT support and some improvements.
DoRA support (and DoHa/DoKr) will be in 2.2.0
1
5
26
FLUX Schnell with TITPOP-200M-dev (the NL+tag prompt gen I'm developing )
I already released the weight but I will not release the introduction until I finish all the things.
(I may provide a close source space for who is interesting to use it)
1
0
25
0
5
24
Just published guanaco-leh-v2!
Larger Dataset
Larger ctx cut off length
BF16 training (not int8)
2
9
25
Next version of kohaku xl is training
Not much difference for well learned concept or style, but have improved detail, feeling, composition and have more data and newer data
Also, natural language caption are used, not only tags.
1
1
25
0
4
23
it's so fucking shit when ppl using the result from 128 x H100 to say "Lora-like have worse detail"
ok, than show me how to use 4x3090 to do native finetune and get same result as 128 x H100?
or give me 128xH100 and I will show you "lora-like" can get almostly same result
5
3
23
This one is better than my previous works.
Plz help this to get higher score than my previous one!!
0
3
21
Some more examples from user.
If you are interested in this model, remember to download it and leave your comments!!!
Imgs (1/?)
The one of best SDXL anime base model: Kohaku-XL delta is released!!!
It is trained on 3.6M images with LoKr algorithm. And it can generate various style and characters
8
79
306
3
2
21
@linaqruf_
Have told you to use "I'm blind" to force gpt4v to give you super detail description
1
0
21