# enginex-ascend-910-transformer-deepseekOCR

运行于【昇腾-910】系列算力卡的【视觉多模态】引擎，基于 transformer 引擎进行架构特别适配优化，支持 DeepSeek-OCR最新开源模型

## QuickStart

1、从 modelscope上下载支持 DeepSeek-OCR
```python
modelscope download --model deepseek-ai/DeepSeek-OCR README.md --local_dir ./model
```
将仓库里的 modeling_deepseekocr.py 复制到模型目录覆盖原本的文件

2、使用Dockerfile生成镜像
从仓库的【软件包】栏目下载基础镜像 git.modelhub.org.cn:9443/enginex-ascend/vllm-ascend:v0.11.0rc0

使用 Dockerfile 生成 镜像
```python
docker build -f Dockerfile -t ascend:deepseek_ocr .
```


3、启动docker
```python
docker run -it --rm \
  -p 10086:80 \
  --name test-ascend-my-1 \
  -v `pwd`:/host \
  -e ASCEND_VISIBLE_DEVICES=1 \
  --device /dev/davinci1:/dev/davinci0 \
  --device /dev/davinci_manager \
  --device /dev/devmm_svm \
  --device /dev/hisi_hdc \
  -v ./model:/model \
  -v /usr/local/dcmi:/usr/local/dcmi \
  -v /usr/local/bin/npu-smi:/usr/local/bin/npu-smi \
  -v /usr/local/Ascend/driver/lib64/:/usr/local/Ascend/driver/lib64/ \
  -v /usr/local/Ascend/driver/version.info:/usr/local/Ascend/driver/version.info \
  -v /etc/ascend_install.info:/etc/ascend_install.info \
  --privileged \
  ascend:deepseek_ocr 
```

4、测试服务
```python
curl -X POST http://localhost:10086/generate \
  -H "Content-Type: application/json" \
  -d '{
    "model": "qwen3-8b",
    "messages": [{"role": "user", "content": "你好"}],
    "stream": true
  }'
```