Banunga_sdxl_lora / README.md
obann001's picture
Update README.md
6503614 verified
|
raw
history blame
6.17 kB
metadata
license: openrail
language:
  - en
metrics:
  - character
pipeline_tag: text-to-image
tags:
  - lora
  - stable-diffusion-xl
  - image-generation
base_model:
  - frankjoshua/realvisxlV40_v40Bakedvae
instance_prompt: Kenza_Né_Banunga
widget:
  - text: >-
      Kenza_Né_Banunga a dark brown skin woman  platinum blonde trapper with
      relaxed mood, crouching, wearing red top, long (black leather coat xl),
      long orange skirt up between bare legs, tattoed forearms, black punk's
      boots, bare fore-arm, marginal, nature,sexy orange panties,
      cameltoeshay_moira_askarah low angle view photo shot hyper realistic of a
      handsome young woman, in a cyberpank style, she is wearing business
      attire, she seems serious and modest
    output:
      url: images/example_2nvir7hcg.png
  - text: >-
      mdjrny-v6 style, Ultra realistic photo portrait of a young woman serious
      Shay Moira Askarah, in a denim jacket standing on the street, looking at
      the camera. cityscape, long, glossy black hair, black eyes, striking
      Persian features, reflecting her Iranian heritage. shaped eyes, subtle
      makeup, high cheekbones and full lips. Her serious expression exudes
      confidence and sophistication. She's holding a black handbag over her
      shoulder. The scene captures the street life, with busy streets and parked
      bike. The image was captured using a DSLR LM-Q710.FGN, 50mm lens, 4k, dslr
      camera, the image has high contrast, highlights on the hair, volumetric
      lighting resulting in a high-resolution photograph that beautifully
      showcases her Persian beauty in Editorial Photography Style --n nsfw
    output:
      url: images/example_86gzu5a5f.png
Generated Image

image/png

  • Model Card for Kenza Né Banunga LoRA SDXL

Example of generated images:

Generated Image Generated Image 1 Generated Image 2 Generated Image 3 Generated Image 4 Generated Image 5 Generated Image 6 Generated Image 7 Generated Image 8 Generated Image 9

Details:

Description:

This model is a Low-Rank Adaptation (LoRA) trained to generate images of a young woman named Kenza Né Banunga. 
It is designed to be used with Stable Diffusion XL models to create photorealistic images of this specific character.

developed_by: Anonymous

finetuned_from: frankjoshua/realvisxlV40_v40Bakedvae

repository: local

  • Trained on Kohya_ss Arch-Linux

Usage:

Direct_use:

This LoRA can be used with compatible SDXL models to generate photorealistic images of Kenza Né Banunga in various situations and poses.

Out_of_scope_use:

This model should not be used to create explicitly sexual, defamatory, or harmful images.

Bias risks limitations:

Description:

The model may have biases in the representation of physical traits and facial expressions. 
The quality of results may vary depending on the base model used.

Recommendations:

Users should be aware of potential biases and use the model responsibly and ethically.

Training details:

Training_data: local

Training_procedure:

Hyperparameters:
  training_regime: bf16 mixed precision
  epochs: 30
  learning_rate: 1e-05
  train_batch_size: 3
  gradient_accumulation_steps: 1
  optimizer: AdamW
  lr_scheduler: constant
  lora_rank: 32
  network_alpha: 128
  resolution: 1024x1024 max
checkpointing:
  save_every_n_steps: 30
  save_at_epoch_end: true

Technical specifications:

Model_architecture:

type: LoRA Standard
base_model: SDXL
gradient_checkpointing: true
xformers: enabled

Compute_infrastructure:

 hardware: NVIDIA GPU GEFORCE RTX-3060
 software: Kohya_ss
  framework: PyTorch with xformers
  precision: bf16 mixed precision

Additional_information:

Training_comment:

3 repeats. More info: https://civitai.com/articles/1771 2_3_Kenza_Né_Banunga woman 35 epoch

Model card:

Authors: Anonymous

Contact: Information not available

Use it with the 🧨 diffusers library

usage example: |

from diffusers import AutoPipelineForText2Image
import torch

pipeline = AutoPipelineForText2Image.from_pretrained(
    'frankjoshua/realvisxlV40_v40Bakedvae', 
    torch_dtype=torch.float16
).to('cuda')
pipeline.load_lora_weights('path/to/Kenza_Né_Banunga_Lora_SDXL_dim32x128', weight_name='pytorch_lora_weights.safetensors')
image = pipeline('your prompt').images

For more details, including weighting, merging, and fusing LoRAs, check the documentation on loading LoRAs in  🧨 diffusers