Welcome to the FFusionAI LyCORIS extracted models repository on Hugging Face & CivitAI! This collection showcases models extracted using quantile extraction mode, as proposed by the LyCORIS framework developed by KohakuBlueleaf
Developed by KohakuBlueleaf, LyCORIS (Lora beYond Conventional methods, Other Rank adaptation Implementations for Stable diffusion) represents the cutting-edge in text-to-image model customization. Refer to the LyCORIS Paper for detailed insights.
Please note that all FFusionAI extracted LyCORIS models are intended for research purposes only and are not licensed for commercial use. We encourage responsible and ethical use of these models to push the boundaries of AI-powered art creation.
π₯ Model: FF.118.Sdxl.Anime.V31.LyCORIS
π₯ Magnitude: 4.9267
π¬ Comment: Leading the pack with the highest UNet magnitude!
π₯ Model: FF.118.Sdxl.Anime.V31.LyCORIS
π₯ Magnitude: 7.5161
π¬ Comment: A standout performer in UNet Conv magnitude!
π₯ Model: FF.103.DynavisionXLAllInOneStylized.Release0557Bakedvae.LyCORIS
π₯ Magnitude: 3.5875
π¬ Comment: Tops in Text Encoder (1) with unmatched magnitude!
Highest Strength in UNet Conv: FF.124.GuchamixXL.V20.LyCORIS with a strength of 0.02023 ποΈ
Most Balanced Model: FF.120.Sdvn6.Realxl.Detailface.LyCORIS, with closely matched magnitudes across all categories βοΈ
High Performer in UNet and UNet Conv: FF.118.Sdxl.Anime.V31.LyCORIS stands out in both UNet and UNet Conv categories π
πͺ Strongest UNet Conv Strength: FF.124.GuchamixXL.V20.LyCORIS (0.02023)
βοΈ Most Balanced Performer: FF.120.Sdvn6.Realxl.Detailface.LyCORIS
π Excelling in maintaining a balanced performance across categories.
π All-Rounder: FF.118.Sdxl.Anime.V31.LyCORIS
π― Dominates in both UNet and UNet Conv categories.
FF.101.juggernautXL_version6Rundiffusion.LyCORIS
FF.102.colossusProjectXLSFW_49bExperimental.LyCORIS
FF.103.dynavisionXLAllInOneStylized_release0557Bakedvae.LyCORIS
FF.104.nightvisionXLPhotorealisticPortrait_release0770Bakedvae.LyCORIS
FF.105.duchaitenAiartSDXL_v20.LyCORIS
FF.106.protovisionXLHighFidelity3D_release0630Bakedvae.LyCORIS
FF.108dynavisionXLAllInOneStylized_release0534bakedvae.LyCORIS
FF.109dreamshaperXL10_alpha2Xl10.LyCORIS
FF.110bluePencilXL_v080.LyCORIS
FF.111bbbSDXL_bbbBetaV2.LyCORIS
FF.112xxmix9realisticsdxl_v10.LyCORIS
FF.113wowXL_v10.LyCORIS
FF.114starlightXLAnimated_v3.LyCORIS
FF.115sdxlYamersCartoonArcadia_v2.LyCORIS
FF.116sdxlUnstableDiffusers_v8HEAVENSWRATH.LyCORIS
FF.117sdxlNijiSpecial_sdxlNijiSE.LyCORIS
FF.118sdxlAnime_v31.LyCORIS
FF.119sdvn7Nijistylexl_v1.LyCORIS
FF.120sdvn6Realxl_detailface.LyCORIS
FF.121samaritan3dCartoon_v40SDXL.LyCORIS
FF.122RealitiesEdgeXL_5.LyCORIS
FF.123LahMysteriousSDXL_v40.LyCORIS
FF.124guchamixXL_v20.LyCORIS
FF.127fantasyStyleBailing_v1007.LyCORIS
FF.118: ###########
(4.93) π
FF.127: ##########
(4.51)
FF.105: ##########
(4.49)
FF.123: ##########
(4.49)
FF.120: #########
(4.47)
FF.118: ###############
(7.52) π
FF.121: #############
(6.87)
FF.127: #############
(6.79)
FF.106: #############
(6.58)
FF.117: #############
(6.58)
FF.103: #######
(3.59) π
FF.110: #######
(3.53)
FF.116: #######
(3.54)
FF.105: #######
(3.55)
FF.111: #######
(3.55)
LyCORIS, standing for "Lora beYond Conventional methods, Other Rank adaptation Implementations for Stable diffusion," offers a new dimension of text-to-image customization. The models here are extracted with a focus on fine-tuning and model evaluation, leveraging techniques such as quantile mode with linear & conv quantile not exceeding 0.2. These extracted models are not direct copies; they encapsulate the essence of the original model, adding a unique twist.
The extraction process for LyCORIS focuses on two key parameters: linear quantile and convolutional quantile. Hereβs a detailed breakdown:
Linear Quantile (--linear_quantile
):
Purpose: Controls the singular value quantile for linear layers in quantile mode.
Setting: Maximum threshold set at 0.2, meaning only singular values up to the 20th percentile are considered.
Impact: Ensures a controlled adaptation from the base model, maintaining subtlety and effectiveness.
Convolutional Quantile (--conv_quantile
):
Purpose: Sets the singular value quantile for convolutional layers.
Setting: Also capped at a maximum of 0.2, aligning with the linear quantile approach.
Impact: Balances the extraction process, ensuring consistency across different types of layers.
π« Policy on Extraction Requests and Algorithm Formats
FFusionAI adheres to specific guidelines regarding extraction requests and the use of different algorithm formats.
1. FF.101.JuggernautXL.Version6Rundiffusion.LyCORIS
UNet Weight Average Magnitude: 4.4698
UNet Weight Average Strength: 0.01224
UNet Conv Weight Average Magnitude: 6.5546
UNet Conv Weight Average Strength: 0.01780
Text Encoder (1) Weight Average Magnitude: 3.5219
Text Encoder (1) Weight Average Strength: 0.01249
2. FF.102.ColossusProjectXLSFW.49bExperimental.LyCORIS
UNet Weight Average Magnitude: 4.3637
UNet Weight Average Strength: 0.01226
UNet Conv Weight Average Magnitude: 6.5158
UNet Conv Weight Average Strength: 0.01751
Text Encoder (1) Weight Average Magnitude: 3.3391
Text Encoder (1) Weight Average Strength: 0.01189
3. FF.103.DynavisionXLAllInOneStylized.Release0557Bakedvae.LyCORIS
UNet Weight Average Magnitude: 4.3940
UNet Weight Average Strength: 0.01285
UNet Conv Weight Average Magnitude: 6.4482
UNet Conv Weight Average Strength: 0.01815
Text Encoder (1) Weight Average Magnitude: 3.5875
Text Encoder (1) Weight Average Strength: 0.01270
4. FF.104.NightvisionXLPhotorealisticPortrait.Release0770Bakedvae.LyCORIS
UNet Weight Average Magnitude: 4.4280
UNet Weight Average Strength: 0.01254
UNet Conv Weight Average Magnitude: 6.5582
UNet Conv Weight Average Strength: 0.01785
Text Encoder (1) Weight Average Magnitude: 3.4918
Text Encoder (1) Weight Average Strength: 0.01286
5. FF.105.DuchaitenAiartSDXL.V20.LyCORIS
UNet Weight Average Magnitude: 4.4919
UNet Weight Average Strength: 0.01255
UNet Conv Weight Average Magnitude: 6.7314
UNet Conv Weight Average Strength: 0.01774
Text Encoder (1) Weight Average Magnitude: 3.5542
Text Encoder (1) Weight Average Strength: 0.01258
6. FF.106.ProtovisionXL.HighFidelity3D.Release0630Bakedvae.LyCORIS
UNet Weight Average Magnitude: 4.4116
UNet Weight Average Strength: 0.01273
UNet Conv Weight Average Magnitude: 6.5832
UNet Conv Weight Average Strength: 0.01780
Text Encoder (1) Weight Average Magnitude: 3.4893
Text Encoder (1) Weight Average Strength: 0.01294
7. FF.108.DynavisionXL.AllInOneStylized.Release0534bakedvae.LyCORIS
UNet Weight Average Magnitude: 4.4738
UNet Weight Average Strength: 0.01247
UNet Conv Weight Average Magnitude: 6.6756
UNet Conv Weight Average Strength: 0.01775
Text Encoder (1) Weight Average Magnitude: 3.5080
Text Encoder (1) Weight Average Strength: 0.01266
8. FF.109.DreamshaperXL.10.Alpha2Xl10.LyCORIS
UNet Weight Average Magnitude: 4.4392
UNet Weight Average Strength: 0.01254
UNet Conv Weight Average Magnitude: 6.5779
UNet Conv Weight Average Strength: 0.01777
Text Encoder (1) Weight Average Magnitude: 3.5054
Text Encoder (1) Weight Average Strength: 0.01236
9. FF.110.BluePencilXL.V080.LyCORIS
UNet Weight Average Magnitude: 4.4079
UNet Weight Average Strength: 0.01273
UNet Conv Weight Average Magnitude: 6.5201
UNet Conv Weight Average Strength: 0.01802
Text Encoder (1) Weight Average Magnitude: 3.5337
Text Encoder (1) Weight Average Strength: 0.01324
10. FF.111.BbbSDXL.BbbBetaV2.LyCORIS
UNet Weight Average Magnitude: 4.3858
UNet Weight Average Strength: 0.01273
UNet Conv Weight Average Magnitude: 6.5305
UNet Conv Weight Average Strength: 0.01789
Text Encoder (1) Weight Average Magnitude: 3.5485
Text Encoder (1) Weight Average Strength: 0.01244
11. FF.112.Xxmix9.Realistic.Sdxl.V10.LyCORIS
UNet Weight Average Magnitude: 4.2945
UNet Weight Average Strength: 0.01352
UNet Conv Weight Average Magnitude: 6.5377
UNet Conv Weight Average Strength: 0.01779
Text Encoder (1) Weight Average Magnitude: 3.4780
Text Encoder (1) Weight Average Strength: 0.01311
12. FF.113.WowXL.V10.LyCORIS
UNet Weight Average Magnitude: 4.4400
UNet Weight Average Strength: 0.01243
UNet Conv Weight Average Magnitude: 6.5624
UNet Conv Weight Average Strength: 0.01787
Text Encoder (1) Weight Average Magnitude: 3.5260
Text Encoder (1) Weight Average Strength: 0.01259
13. FF.114.StarlightXL.Animated.V3.LyCORIS
UNet Weight Average Magnitude: 4.3787
UNet Weight Average Strength: 0.01278
UNet Conv Weight Average Magnitude: 6.5010
UNet Conv Weight Average Strength: 0.01788
Text Encoder (1) Weight Average Magnitude: 3.5063
Text Encoder (1) Weight Average Strength: 0.01257
14. FF.115.Sdxl.Yamers.Cartoon.Arcadia.V2.LyCORIS
UNet Weight Average Magnitude: 4.3740
UNet Weight Average Strength: 0.01235
UNet Conv Weight Average Magnitude: 6.5276
UNet Conv Weight Average Strength: 0.01787
Text Encoder (1) Weight Average Magnitude: 3.5094
Text Encoder (1) Weight Average Strength: 0.01239
15. FF.116.Sdxl.Unstable.Diffusers.V8.HeavensWrath.LyCORIS
UNet Weight Average Magnitude: 4.3866
UNet Weight Average Strength: 0.01239
UNet Conv Weight Average Magnitude: 6.5490
UNet Conv Weight Average Strength: 0.01778
Text Encoder (1) Weight Average Magnitude: 3.5354
Text Encoder (1) Weight Average Strength: 0.01238
16. FF.117.Sdxl.Niji.Special.Sdxl.NijiSE.LyCORIS
UNet Weight Average Magnitude: 4.4737
UNet Weight Average Strength: 0.01221
UNet Conv Weight Average Magnitude: 6.5781
UNet Conv Weight Average Strength: 0.01777
Text Encoder (1) Weight Average Magnitude: 3.5071
Text Encoder (1) Weight Average Strength: 0.01236
17. FF.118.Sdxl.Anime.V31.LyCORIS
UNet Weight Average Magnitude: 4.9267
UNet Weight Average Strength: 0.01515
UNet Conv Weight Average Magnitude: 7.5161
UNet Conv Weight Average Strength: 0.01696
Text Encoder (1) Weight Average Magnitude: 3.5106
Text Encoder (1) Weight Average Strength: 0.01236
18. FF.119.Sdvn7.Nijistylexl.V1.LyCORIS
UNet Weight Average Magnitude: 4.4627
UNet Weight Average Strength: 0.01241
UNet Conv Weight Average Magnitude: 6.5621
UNet Conv Weight Average Strength: 0.01778
Text Encoder (1) Weight Average Magnitude: 3.5169
Text Encoder (1) Weight Average Strength: 0.01240
19. FF.120.Sdvn6.Realxl.Detailface.LyCORIS
UNet Weight Average Magnitude: 4.4728
UNet Weight Average Strength: 0.01220
UNet Conv Weight Average Magnitude: 6.5779
UNet Conv Weight Average Strength: 0.01775
Text Encoder (1) Weight Average Magnitude: 3.5054
Text Encoder (1) Weight Average Strength: 0.01236
20. FF.121.Samaritan3d.Cartoon.V40SDXL.LyCORIS
UNet Weight Average Magnitude: 4.5336
UNet Weight Average Strength: 0.01255
UNet Conv Weight Average Magnitude: 6.8741
UNet Conv Weight Average Strength: 0.01782
Text Encoder (1) Weight Average Magnitude: 3.5074
Text Encoder (1) Weight Average Strength: 0.01236
21. FF.122.RealitiesEdgeXL.5.LyCORIS
UNet Weight Average Magnitude: 4.3591
UNet Weight Average Strength: 0.01286
UNet Conv Weight Average Magnitude: 6.3116
UNet Conv Weight Average Strength: 0.01835
Text Encoder (1) Weight Average Magnitude: 3.5195
Text Encoder (1) Weight Average Strength: 0.01267
22. FF.123.LahMysteriousSDXL.V40.LyCORIS
UNet Weight Average Magnitude: 4.4911
UNet Weight Average Strength: 0.01236
UNet Conv Weight Average Magnitude: 6.6664
UNet Conv Weight Average Strength: 0.01778
Text Encoder (1) Weight Average Magnitude: 3.5054
Text Encoder (1) Weight Average Strength: 0.01236
23. FF.124.GuchamixXL.V20.LyCORIS
UNet Weight Average Magnitude: 4.3563
UNet Weight Average Strength: 0.01375
UNet Conv Weight Average Magnitude: 5.6046
UNet Conv Weight Average Strength: 0.02023
Text Encoder (1) Weight Average Magnitude: 3.5105
Text Encoder (1) Weight Average Strength: 0.01236
24. FF.127.FantasyStyleBailing.V1007.LyCORIS
UNet Weight Average Magnitude: 4.5056
UNet Weight Average Strength: 0.01246
UNet Conv Weight Average Magnitude: 6.7862
UNet Conv Weight Average Strength: 0.01772
Text Encoder (1) Weight Average Magnitude: 3.5054
Text Encoder (1) Weight Average Strength: 0.01236
Higher Quantile Values:
π Policy: Quantile values for extractions are capped at 0.2.
π Reason: To preserve the base model's essence and stability.
No Extraction of Other People's Work:
β οΈ Policy: Prohibits extracting models or works created by others.
π Reason: To respect intellectual property and prevent unauthorized use.
No Adaptation to Different Algorithm Formats:
β Policy: We do not accept requests to adapt or convert models to other formats without the proper approval of the weights' owner.
These structured guidelines are crucial for maintaining the quality, legality, and ethical standards of the LyCORIS repository, fostering responsible and effective use.
Before accessing or using the models, please review the full license agreement.
The correct license and permissions can be found at:
"Model Weights: The weights for the LyCORIS models are provided "as is." FFusion AI and Source Code Bulgaria do not grant any rights for their commercial use. These weights are strictly for testing and experimental purposes.
ORIGIN OF LYCORIS:
The LyCORIS models and weights provided are extracted based on the FFusion/FFusionXL-BASE model. All licenses, terms, and conditions set by the original creators must be adhered to."
Research and Testing Focus: The models and weights available in this repository are strictly for research and testing purposes. They are not generally intended for commercial use and are subject to the conditions of each individual LyCORIS extraction.
Content Disclaimer: FFusion AI, Source Code Bulgaria Ltd, and BlackswanTechnologies do not endorse or guarantee the content produced by the weights in each LyCORIS model. There's potential for generating NSFW or offensive content. Collectively, we expressly disclaim responsibility for the outcomes and content produced by these weights.
Model Acknowledgement: The FFusionXL-BASE model is a unique development by FFusion AI. Rights to this and associated modifications belong to FFusion AI and Source Code Bulgaria Ltd. Users must adhere to both this license and any conditions set by Stability AI Ltd for referenced models.
For Non-Commercial Use: Models are primarily for research and experimental use.
Content Disclaimer: Potential for NSFW or offensive content generation.
@misc{LyCORIS,
title={Navigating Text-To-Image Customization: From LyCORIS Fine-Tuning to Model Evaluation},
author={Shin-Ying Yeh and Yu-Guan Hsieh and Zhidong Gao and Bernard B W Yang and Giyeong Oh and Yanmin Gong},
year={2023},
eprint={2309.14859},
archivePrefix={arXiv}
}
π FFusion.ai Contact Information
Proudly maintained by Source Code Bulgaria Ltd & Black Swan Technologies.
π§ For collaborations, inquiries, or support: [email protected]
π Locations: Sofia | Istanbul | London
Connect with Us:
Our Websites:
π FFusion.ai
π FFAI.eu
π 1e-2.com