noobai X L V Pred cyberfixv2 perpendicularcyberfixv2
PanchovixNOOBAI-XL-VPRED-CYBERFIXV2-PERPENDICULARCYBERFIXV2
Introduction
The NOOBAI-XL-VPRED-CYBERFIXV2-PERPENDICULARCYBERFIXV2 is an advanced text-to-image generative model based on Laxhar's NOOBAI-XL. It utilizes comprehensive datasets such as Danbooru and e621, integrating native tags with natural language captioning. This model operates distinctly as a v-prediction model and requires specific configurations for optimal performance.
Architecture
- Developed by: Laxhar Lab
- Model Type: Diffusion-based text-to-image generative model
- Fine-tuned from: Laxhar/noobai-XL_v1.0
- Sponsored by: Lanyun Cloud
Training
The model is trained using the latest Danbooru and e621 datasets, ensuring high-quality image generation. It employs data normalization, time-based decay coefficients, and ranks images to enhance quality tag accuracy. The model is fine-tuned for specific quality and aesthetic tags, considering user preferences and image aesthetics.
Guide: Running Locally
Method I: ReForge
- Install ReForge by following the instructions here.
- Launch WebUI and use the model.
Method II: ComfyUI
- Install ComfyUI by following the instructions here.
- Use the model with sample nodes.
Method III: WebUI
- Install WebUI by following the instructions here.
- Switch to the development branch using
git switch dev
. - Pull the latest updates using
git pull
. - Launch WebUI and use the model.
Method IV: Diffusers
import torch
from diffusers import StableDiffusionXLPipeline, EulerDiscreteScheduler
ckpt_path = "/path/to/model.safetensors"
pipe = StableDiffusionXLPipeline.from_single_file(
ckpt_path,
use_safetensors=True,
torch_dtype=torch.float16,
)
scheduler_args = {"prediction_type": "v_prediction", "rescale_betas_zero_snr": True}
pipe.scheduler = EulerDiscreteScheduler.from_config(pipe.scheduler.config, **scheduler_args)
pipe.enable_xformers_memory_efficient_attention()
pipe = pipe.to("cuda")
prompt = "masterpiece, best quality, ..."
negative_prompt = "nsfw, worst quality, ..."
image = pipe(
prompt=prompt,
negative_prompt=negative_prompt,
width=832,
height=1216,
num_inference_steps=28,
guidance_scale=5,
generator=torch.Generator().manual_seed(42),
).images[0]
image.save("output.png")
Recommended Settings
- CFG: 4 ~ 5
- Steps: 28 ~ 35
- Sampling Method: Euler
- Resolution: Around 1024x1024, with specific options like 768x1344, 832x1216, etc.
Cloud GPUs
For better performance, consider using cloud GPUs from providers like AWS, Google Cloud, or Azure.
License
The model is licensed under the fair-ai-public-license-1.0-sd, with specific usage restrictions:
- Usage Restrictions: Prohibits harmful, malicious, or illegal activities and the generation of unethical content.
- Commercial Prohibition: Commercial use, monetization, or derivative commercial products are prohibited.
- Open Source Community: Requires open source derivative works and shared synthesis details.
- Disclaimer: Users must assume risks and potential consequences of usage.