SDXL-ProteusSigma Training with ZTSNR and NovelAI V3 Improvements

- [x] 10k dataset proof of concept (completed)[link](https://huggingface.co/dataautogpt3/ProteusSigma)

- [x] 500k+ dataset finetune (completed) [Note: not aesthetically tuned whatsoever]

- [ ] 12M million dataset finetune (planned)


Proteus and Mobius datasets

CUSTOM INFERENCE IS REQUIRED FOR BEST RESULTS!

https://github.com/DataCTE/SDXL-Training-Improvements/tree/main/Comfyui-zsnrnode

use this comfyui custom node from the training repo.

and the workflow here: https://github.com/DataCTE/SDXL-Training-Improvements/blob/main/Comfyui-zsnrnode/ztsnr%2Bv-pred.json

Model Details

• Model Type: SDXL Fine-tuned with ZTSNR and NovelAI V3 Improvements

• Base Model: stabilityai/stable-diffusion-xl-base-1.0

• Training Dataset: 500,000 high-quality images

• License: Apache 2.0

Key Features

• Zero Terminal SNR (ZTSNR) implementation

• Increased σ_max ≈ 20000.0 (NovelAI research)

Training Details

Training Configuration

• Learning Rate: 4e-7

• Batch Size: 8

• Gradient Accumulation Steps: 8

• Optimizer: AdamW

• Precision: bfloat16

• Epochs: 80

Performance Improvements

• 47% fewer artifacts at σ < 5.0

• Stable composition at σ > 12.4

• 31% better detail consistency

• Improved color accuracy

• Enhanced dark tone reproduction

Repository and Resources

• GitHub Repository: SDXL-Training-Improvements

• Training Code: Available in the repository

• Documentation: Implementation Details

• Issues and Support: GitHub Issues

Citation

@article{ossa2024improvements,
  title={Improvements to SDXL in NovelAI Diffusion V3},
  author={Ossa, Juan and Doğan, Eren and Birch, Alex and Johnson, F.},
  journal={arXiv preprint arXiv:2409.15997v2},
  year={2024}
}