File size: 1,171 Bytes
2f12cef
 
 
8719c3e
2f12cef
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c4c1060
2f12cef
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
63d9864
c767ef4
2f12cef
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
---
license: mit
library_name: rkllm
base_model: deepseek-ai/DeepSeek-OCR
tags:
  - rkllm
  - rk3588
  - npu
  - rockchip
  - quantized
  - edge-ai
  - orange-pi
  - vision-language
  - ocr
model_name: DeepSeek-OCR-RKLLM
pipeline_tag: image-text-to-text
language:
  - en
---


# DeepSeek-OCR for RK3588

This repository contains the optimized **RKNN** (Vision) and **RKLLM** (Language) models for running **DeepSeek-OCR** on Rockchip RK3588 NPUs.

### 📦 Files

| File | Type | Description |
| --- | --- | --- |
| `deepseekocr_vision_rk3588.rknn` | **RKNN** | Vision Encoder (ViT) |
| `deepseekocr_w8a8_rk3588.rkllm` | **RKLLM** | Language Decoder (w8a8) |

---

### ⚙️  System Requirements

* **NPU Driver:** `≥ 0.9.8`
* **RKLLM Runtime:** `v1.2.3`
* **RKNN-Toolkit2:** `v2.3.0`

### 💻 Tested Hardware

* **Device:** Orange Pi 5 Pro
* **RAM:** 16GB
* **OS:** Armbian OS

---

### 🙏 Acknowledgements

* [DeepSeek Team](https://huggingface.co/deepseek-ai) for the base model architecture.
* [Rockchip / airockchip](https://github.com/airockchip/rknn-llm) for the RKLLM toolkit and runtime.
* Converted by [Akmal Dira](https://huggingface.co/akmaldira).

---