————————————————————————————————————————

AiARTiST XL Basic Unit CADS2 LoRA Compatible Version SDXL Large Model Hyper Edition

AiARTiST XL Basic Unit CADS2 LoRA Compatible Version_HyperSD_STEPS8_CFG1.safetensors

----------------------------------------------------------------------

AiARTiST XL UNIT-lib Basic Unit Hyper-SDXL Accelerated Version

Built-in Accelerator Accelerator: Hyper-SDXL-8steps-lora.safetensors

Hyper-SD is one of the latest diffusion model acceleration technologies, lossless HD acceleration, suitable for CN adaptation

After testing on multiple base models, Hyper-SD can derive more image information than before acceleration!

Strongly recommend using Hyper version for 8G-16G memory and MAC users, direct output of high-speed 2048 HD images

----------------------------------------------------------------------

Parameter configuration for CADS-HyperRAY XL: (Important)

Clip Skip 1 / CFG 1 / Steps 10 / Sampling Method (Sampler) Euler a / VAE Auto

Size: 1280x720 / 1536x768 / 960x1920 / 1920x960 / 2048x1024 / 1024x2048

For sizes larger than 1536, it is recommended to open the Kohya HRFix plugin to correct composition errors and enlarge the corrections

Avoid using negative words, no need to fill in negative words, well-adapted for LoRA and ControlNet.

----------------------------------------------------------------------

AiARTiST Basic Unit CADS2 LoRA Compatible Version SDXL Large Model Metadata Information Disclosure

File Name: AiARTiST_UNIT_LiberATION_HyperSD.safetensors

{

   "ckpt_name1": "SD XL\\AiARTiST CADS v1.0 experienceflow",

   "ckpt_name2": "SD XL\\MOHAWK v2.0",

   "ckpt_name3": "SD XL\\Jib Mix Realistic XL v8.0",

   "ckpt_name4": "SD XL\\LEOSAM's HelloWorld XL 5.0",

   "ckpt_name5": "SD XL\\Hephaistos NextGen DPO",

   "ckpt_name6": "SD XL\\Proteus v0.2",

   "ckpt_name7": "SD XL\\Aetherverse XL v1.0",

   "ckpt_name8": "SD XL\\AiARTiST CADS v2.0 justice_alliance",

   "LoRA_merge1": "SD XL\\AiARTiST-1GIRL",

   "LoRA_merge2": "SD XL\\Hyper-SDXL-8steps",

   "Training weights": "SD3-Preview"

}

----------------------------------------------------------------------

Q&A

Is VAE built-in? Built-in SDXL-FP16-fixt VAE

----------------------------------------------------------------------

Appendix: CreativeML Open RAIL-M Introduction

The CreativeML Open RAIL-M license was established by Robin Rombach, Patrick Esser, and their contributors on August 22, 2022, to standardize the use, replication, and distribution behavior of multi-modal generation models, and to promote open and responsible downstream use of models. This license draws on the spirit of open source licenses, while incorporating restriction clauses based on usage scenarios to prevent potential misuse of models.

The license covers contents including models and their derivatives, defined as follows:

1. "License" refers to the terms and conditions of use, reproduction, and distribution of models and their derivatives as defined in this document.
2. "Data" refers to information and content in the dataset used for training, pre-training, or evaluating the models, but data is not within the scope of this license.
3. "Output" refers to the result information content generated by operating the model.
4. "Model" refers to a collection based on machine learning, including trained weights, parameters (including optimizer state), trained or adjusted according to the model architecture in the accompanying materials, at least partially trained based on data.
5. "Model Derivatives" refers to all modifications to the model, works based on the model, and other models that initialize patterns based on the weights, parameters, activations, or outputs of the model to make another model perform similarly, including but not limited to distillation methods and other similarity transfer methods.
6. "Accompanying Materials" refer to the source code and scripts used to define, run, load, and evaluate the models, as well as related documents, tutorials, and examples for preparing training or evaluation data, etc.
7. "Distribution" refers to the transmission, reproduction, publication, or sharing of models or their derivatives with third parties, including hosting services provided in API or web access forms.

The license holder (copyright owner or authorized entity) grants users a perpetual, global, non-exclusive, royalty-free copyright and patent license to copy, display, perform, distribute, modify the models and their accompanying materials and derivatives, but must adhere to usage restrictions (see Appendix A).

When distributing models or their derivatives, users must include the same usage restrictions and ensure that the recipients are aware of these constraints. Users are responsible for the outputs generated by their use of the model and must ensure that they comply with the license provisions.

Additionally, the license holder reserves the right, as required by law, to restrict the misuse of models, update models electronically, or adjust model outputs according to updates.

Overall, the CreativeML Open RAIL-M license aims to balance open and responsible AI development by guiding the responsible use of models in open scientific practices in art creation and other fields through clear terms.

————————————————————————————————————————

For testing questions, please leave a message, for business cooperation, visit the personal homepage +V Zyuan980

AiARTiST | Metaverse China·Shandong Digital Human/AiGC/Metaverse Direction

Being a good tool personnel serving artists Contact method:https://afdian.net/a/AiARTiST-lib

————————————————————————————————————————