36 lines
1001 B
Markdown
36 lines
1001 B
Markdown
|
# enginex-r200-vc-cnn
|
|||
|
|
|
|||
|
|
运行于昆仑芯 R200 算力卡的【视觉分类】引擎，基于 CNN 架构，支持 BEiT、MobileViT 等流行模型
|
|||
|
|
|
|||
|
|
## QuickStart
|
|||
|
|
|
|||
|
|
1、从 modelscope上下载视觉分类的模型，例如 microsoft/beit-base-patch16-224
|
|||
|
|
```python
|
|||
|
|
modelscope download --model microsoft/beit-base-patch16-224
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
2、使用Dockerfile生成镜像
|
|||
|
|
使用 Dockerfile 生成 镜像
|
|||
|
|
```python
|
|||
|
|
docker build -f Dockerfile_kunlunxin -t kunlunxin-my:v1 .
|
|||
|
|
```
|
|||
|
|
其中基础镜像 r200-8f_xmlir-ubuntu_2004_x86_64:v0.27 通过联系昆仑芯厂商技术支持可获取
|
|||
|
|
注意 Dockerfile 中已预先将模型 microsoft_beit_base_patch16_224_pt22k_ft22k 放在了 /model 下面
|
|||
|
|
|
|||
|
|
3、启动docker
|
|||
|
|
```python
|
|||
|
|
docker run -it --rm \
|
|||
|
|
--privileged \
|
|||
|
|
-p 10086:80 \
|
|||
|
|
--device /dev/xpu0 \
|
|||
|
|
--device /dev/xpuctrl \
|
|||
|
|
--name kunlun-my-v4 \
|
|||
|
|
kunlunxin-my:v1
|
|||
|
|
```
|
|||
|
|
|
|||
|
|
4、测试服务
|
|||
|
|
```python
|
|||
|
|
curl -X POST http://localhost:10086/v1/private/s782b4996 \
|
|||
|
|
> -F "image=@/home/zhoushasha/models/026_0010.jpg"
|
|||
|
|
```
|