Create install_npu.md by light-lee-2016 · Pull Request #226 · Yuliang-Liu/MonkeyOCR (original) (raw)

@@ -0,0 +1,57 @@

# Install with NPU Support

此向导仅用于在NPU上运行monkey-ocr，需提前下载模型权重到/root/.cache目录，已在910B4显卡上面做过验证，后端使用vllm_queue, 性能大致0.7page/s。

# step1: 下载vllm-ascend镜像并启动，需要vllm-ascend:v0.9.2rc1及更高版本

```

# Update DEVICE according to your device (/dev/davinci[0-7])

export DEVICE=/dev/davinci0

# Update the vllm-ascend image

export IMAGE=quay.io/ascend/vllm-ascend:v0.9.2rc1

docker run --rm \

--name vllm-ascend \

--device $DEVICE \

--device /dev/davinci_manager \

--device /dev/devmm_svm \

--device /dev/hisi_hdc \

-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 \

-v /root/.cache:/root/.cache \

-p 8000:8000 \

-it $IMAGE bash

```

# step2: 进入镜像内部安装对应的包

```

git clone https://github.com/Yuliang-Liu/MonkeyOCR.git

cd MonkeyOCR

pip install -e .

# 下面的命令不执行会报错缺少一些opencv图形化显示的so，我们不需要

pip uninstall opencv-python-headless opencv-python -y

pip install opencv-python-headless==4.11.0.86

```

# step3:修改model_config.yaml配置

```

device: npu

weights:

doclayout_yolo: Structure/doclayout_yolo_docstructbench_imgsz1280_2501.pt # or Structure/layout_zh.pt

# PP-DocLayout_plus-L: Structure/PP-DocLayout_plus-L

layoutreader: Relation

layout_config:

model: doclayout_yolo # PP-DocLayout_plus-L (MonkeyOCR-pro, npu不支持) / doclayout_yolo (MonkeyOCR)

reader:

name: layoutreader

models_dir: /root/.cache

chat_config:

weight_path: /root/.cache/Recognition

backend: vllm_queue

...

```

# step4: 启动服务

```

export VLLM_WORKER_MULTIPROC_METHOD=spawn

export VLLM_USE_V1=1

# 可选，如果是卡0可以不设置

export ASCEND_RT_VISIBLE_DEVICEAS=0

univcore api.main:app --host 0.0.0.0 --port 8000

```