---
license: creativeml-openrail-m
tags:
- stable-diffusion
- stable-diffusion-diffusers
- text-to-image
datasets:
- ChristophSchuhmann/improved_aesthetics_6.25plus
library_name: diffusers
pipeline_tag: text-to-image
extra_gated_prompt: >-
  This model is open access and available to all, with a CreativeML OpenRAIL-M
  license further specifying rights and usage.

  The CreativeML OpenRAIL License specifies: 


  1. You can't use the model to deliberately produce nor share illegal or
  harmful outputs or content 

  2. The authors claim no rights on the outputs you generate, you are free to
  use them and are accountable for their use which must not go against the
  provisions set in the license

  3. You may re-distribute the weights and use the model commercially and/or as
  a service. If you do, please be aware you have to include the same use
  restrictions as the ones in the license and share a copy of the CreativeML
  OpenRAIL-M to all your users (please read the license entirely and carefully)

  Please read the full license carefully here:
  https://huggingface.co/spaces/CompVis/stable-diffusion-license
      
extra_gated_heading: Please read the LICENSE to access this model

---


# BK-SDM-2M Model Card

BK-SDM-{[**Base-2M**](https://huggingface.co/nota-ai/bk-sdm-base-2m), [**Small-2M**](https://huggingface.co/nota-ai/bk-sdm-small-2m), [**Tiny-2M**](https://huggingface.co/nota-ai/bk-sdm-tiny-2m)} are pretrained with **10× more data** (2.3M LAION image-text pairs) compared to our previous release.
- Block-removed Knowledge-distilled Stable Diffusion Model (BK-SDM) is an architecturally compressed SDM for efficient text-to-image synthesis. 
- The previous BK-SDM-{[Base](https://huggingface.co/nota-ai/bk-sdm-base), [Small](https://huggingface.co/nota-ai/bk-sdm-small), [Tiny](https://huggingface.co/nota-ai/bk-sdm-tiny)} were obtained via distillation pretraining on 0.22M LAION pairs.
- Resources for more information: [Paper](https://arxiv.org/abs/2305.15798), [GitHub](https://github.com/Nota-NetsPresso/BK-SDM), [Demo]( https://huggingface.co/spaces/nota-ai/compressed-stable-diffusion).


## Examples with 🤗[Diffusers library](https://github.com/huggingface/diffusers).

An inference code with the default PNDM scheduler and 50 denoising steps is as follows.

```python
import torch
from diffusers import StableDiffusionPipeline

pipe = StableDiffusionPipeline.from_pretrained("nota-ai/bk-sdm-tiny-2m", torch_dtype=torch.float16)
pipe = pipe.to("cuda")

prompt = "a black vase holding a bouquet of roses"
image = pipe(prompt).images[0]  
    
image.save("example.png")
```


## Compression Method 

Adhering to the [U-Net architecture](https://huggingface.co/nota-ai/bk-sdm-tiny#u-net-architecture) and [distillation pretraining](https://huggingface.co/nota-ai/bk-sdm-tiny#distillation-pretraining) of BK-SDM, the difference in BK-SDM-2M is a 10× increase in the number of training pairs.

- **Training Data**: 2,256,472 image-text pairs (i.e., 2.3M pairs) from [LAION-Aesthetics V2 6.25+](https://laion.ai/blog/laion-aesthetics/).
- **Hardware:** A single NVIDIA A100 80GB GPU
- **Gradient Accumulations**: 4
- **Batch:** 256 (=4×64)
- **Optimizer:** AdamW
- **Learning Rate:** a constant learning rate of 5e-5 for 50K-iteration pretraining


## Experimental Results 

The following table shows the zero-shot results on 30K samples from the MS-COCO validation split. After generating 512×512 images with the PNDM scheduler and 25 denoising steps, we downsampled them to 256×256 for evaluating generation scores.

- Our models were drawn at the 50K-th training iteration.

| Model | FID↓ | IS↑ | CLIP Score↑<br>(ViT-g/14) | # Params,<br>U-Net | # Params,<br>Whole SDM |
|---|:---:|:---:|:---:|:---:|:---:|
| [Stable Diffusion v1.4](https://huggingface.co/CompVis/stable-diffusion-v1-4) | 13.05 | 36.76 | 0.2958 | 0.86B | 1.04B |
| [BK-SDM-Base](https://huggingface.co/nota-ai/bk-sdm-base) (Ours) | 15.76 | 33.79 | 0.2878 | 0.58B | 0.76B |
| [BK-SDM-Base-2M](https://huggingface.co/nota-ai/bk-sdm-base-2m) (Ours) | 14.81 | 34.17 | 0.2883 | 0.58B | 0.76B |
| [BK-SDM-Small](https://huggingface.co/nota-ai/bk-sdm-small) (Ours) | 16.98 | 31.68 | 0.2677 | 0.49B | 0.66B |
| [BK-SDM-Small-2M](https://huggingface.co/nota-ai/bk-sdm-small-2m) (Ours) | 17.05 | 33.10 | 0.2734 | 0.49B | 0.66B |
| [BK-SDM-Tiny](https://huggingface.co/nota-ai/bk-sdm-tiny) (Ours) | 17.12 | 30.09 | 0.2653 | 0.33B | 0.50B |
| [BK-SDM-Tiny-2M](https://huggingface.co/nota-ai/bk-sdm-tiny-2m) (Ours) | 17.53 | 31.32 | 0.2690 | 0.33B | 0.50B |



### Effect of Different Data Sizes for Training BK-SDM-Small
Increasing the number of training pairs improves the IS and CLIP scores over training progress. The MS-COCO 256×256 30K benchmark was used for evaluation.

<center>
    <img alt="Training progress with different data sizes" img src="https://netspresso-research-code-release.s3.us-east-2.amazonaws.com/assets-bk-sdm/fig_iter_data_size.png" width="100%">
</center>

Furthermore, with the growth in data volume, visual results become more favorable (e.g., better image-text alignment and clear distinction among objects). 

<center>
    <img alt="Visual results with different data sizes" img src="https://netspresso-research-code-release.s3.us-east-2.amazonaws.com/assets-bk-sdm/fig_results_data_size.png" width="100%">
</center>

### Additional Visual Examples

<center>
    <img alt="additional visual examples" img src="https://netspresso-research-code-release.s3.us-east-2.amazonaws.com/assets-bk-sdm/fig_results_models_2m.png" width="100%">
</center>


# Uses 
Follow [the usage guidelines of Stable Diffusion v1](https://huggingface.co/CompVis/stable-diffusion-v1-4#uses). 


# Acknowledgments
- We express our gratitude to [Microsoft for Startups Founders Hub](https://www.microsoft.com/en-us/startups) for generously providing the Azure credits used during pretraining.
- We deeply appreciate the pioneering research on Latent/Stable Diffusion conducted by [CompVis](https://github.com/CompVis/latent-diffusion), [Runway](https://runwayml.com/), and [Stability AI](https://stability.ai/).
- Special thanks to the contributors to [LAION](https://laion.ai/), [Diffusers](https://github.com/huggingface/diffusers), and [Gradio](https://www.gradio.app/) for their valuable support.


# Citation
```bibtex
@article{kim2023architectural,
  title={On Architectural Compression of Text-to-Image Diffusion Models},
  author={Kim, Bo-Kyeong and Song, Hyoung-Kyu and Castells, Thibault and Choi, Shinkook},
  journal={arXiv preprint arXiv:2305.15798},
  year={2023},
  url={https://arxiv.org/abs/2305.15798}
}
```
```bibtex
@article{Kim_2023_ICMLW,
  title={BK-SDM: Architecturally Compressed Stable Diffusion for Efficient Text-to-Image Generation},
  author={Kim, Bo-Kyeong and Song, Hyoung-Kyu and Castells, Thibault and Choi, Shinkook},
  journal={ICML Workshop on Efficient Systems for Foundation Models (ES-FoMo)},
  year={2023},
  url={https://openreview.net/forum?id=bOVydU0XKC}
}
```

*This model card was written by Bo-Kyeong Kim and is based on the [Stable Diffusion v1 model card]( https://huggingface.co/CompVis/stable-diffusion-v1-4).*