---
license: apache-2.0
---

# SD1.5_rknn_3588_euler (Assets)

This repository contains **pre-converted binary assets** for running **Stable Diffusion 1.5 (Euler)** on **Rockchip RK3588 / RK3588S** using **RKNN**.

⚠️ **Important**  
This is **NOT** a Hugging Face “model” intended for `from_pretrained()`.  
It is a **DATASET** that stores hardware-specific binaries (`.rknn`) and weights.

The runtime, CLI, and WebUI are hosted separately on GitHub.

---

## What this repository contains

.
├── models/
│ └── business_rknn/
│ ├── unet/ # SD1.5 UNet (Realistic Vision–based)
│ │ └── model.rknn
│ ├── text_encoder/
│ ├── vae_encoder/
│ ├── vae_decoder/
│ └── scheduler/
│
├── model/
│ └── sr/
│ └── realesrgan/
│ └── realesrgan_x4plus_tile128_fp16.rknn
│
└── gfpgan/
└── weights/
├── GFPGANv1.4.pth
├── detection_Resnet50_Final.pth
└── parsing_parsenet.pth



---

## Model details

- **Base architecture:** Stable Diffusion 1.5  
- **Primary checkpoint:** Realistic Vision (SD1.5)
- **Scheduler:** Euler
- **Runtime format:** RKNN
- **Target hardware:** Rockchip RK3588 / RK3588S

---

## Super-Resolution

- **Real-ESRGAN x4 (RKNN)**
- ⚠️ Upscaling is **limited to ×4 only**  
  (this is a model limitation, not configurable)

---

## Intended usage

These assets are meant to be **downloaded explicitly** by a custom runtime.

Typical usage pattern:
- Clone runtime code from GitHub
- Download this dataset via `huggingface_hub.snapshot_download()`
- Place files into expected directories
- Run inference using RKNN

➡️ Runtime & usage instructions:  
**GitHub:** https://github.com/Mojo24x7/SD1.5_rknn_3588_euler

---

## License & credits

- All model weights are provided under their **original upstream licenses**
- This repository does **not claim ownership** of:
  - Stable Diffusion
  - Realistic Vision
  - Real-ESRGAN
  - GFPGAN

See the GitHub repository for full credits and acknowledgements.

---

## Disclaimer

This dataset is provided for **research, experimentation, and personal use**.  
The author is not affiliated with Stability AI, CompVis, Rockchip, or Hugging Face.