---
license: apache-2.0
---

# DSOCR
**Enhanced Inference API and User Interface for Using DeepSeek OCR Model on Windows/Mac and so on** 

增强的DeepSeek OCR推理 API 和用户界面，用于在 Windows/Mac 等系统上使用 DeepSeek OCR 模型



# Features 特性

* **Efficient inference of DeepSeek-OCR models using multiple devices (CUDA/CPU) at multiple precisions (fp16/bf16/int8/nf4)**
* 使用多种设备(CUDA/CPU)在多种精度下(fp16/bf16/int8/nf4)高效推理DeepSeek-OCR模型
* **Exclusive custom inference API optimizes the speed, performance, and display of the model inference pipeline and supports different operating systems**
* 独家自定义推理API，优化了模型推理管线的速度、性能、显示，并支持不同操作系统
* **It has a modern, simple, easy-to-use and user-friendly reasoning GUI. All operations are fully automatic and no code is required**
* 具有现代、简洁、易用、人性化的推理GUI，一切操作完全自动化、无需任何代码
* **It supports working image cloning/importing and supports multi-task reasoning (in the future), allowing you to save or restore your complex workflow at any time**
* 它支持工作镜像克隆/导入，并支持多任务推理(未来)，使得您可以随时保存或还原您的复杂工作流
* **Support running as a server and running inference requests using standard http/https protocols (in the future)**
* 支持作为服务器运行，使用标准http/https协议运行推理请求(未来)


# Python Version Recommendation 推荐的Python版本

* `Python 3.10.x`
* `Python 3.12.x`

* `pip command`
```
pip install torch==2.9.1 torchvision==0.24.0 transformers==4.46.3 tokenizers==0.20.3 bitsandbytes>=0.41.0 accelerate>=0.26.0 tqdm einops addict easydict PyQt6 zstandard numpy pandas Pillow pycryptodome pymupdf img2pdf reportlab
```


# Version History 版本历史

* V0 build 0.0.1 - 2025-10-28


# GUI Preview 图形界面预览

<img width="1476" height="861" alt="image" src="https://github.com/user-attachments/assets/8a2e1ad7-3658-4e5e-980c-35078791654f" />


`GUI by Nathmath/Katz/dof-studio`