能否提供一个在910b上进行pp-ocr v4部署的样例？

[bltcn]

运行环境为： EulerOS 2.0 (SP8), CANN-6.0.1 +------------------------------------------------------------------------------------------------+ | npu-smi 23.0.rc2.2 Version: 23.0.rc2.2 | +---------------------------+---------------+----------------------------------------------------+ | NPU Name | Health | Power(W) Temp(C) Hugepages-Usage(page)| | Chip | Bus-Id | AICore(%) Memory-Usage(MB) HBM-Usage(MB) | +===========================+===============+====================================================+ | 2 910B1 | OK | 96.2 41 0 / 0 | | 0 | 0000:C2:00.0 | 0 0 / 0 4349 / 65536 | +===========================+===============+====================================================+ +---------------------------+---------------+----------------------------------------------------+ | NPU Chip | Process id | Process name | Process memory(MB) | +===========================+===============+====================================================+ | No running processes found in NPU 2 | +===========================+===============+====================================================+

[horcham]

你好，910b机器可以使用如下步骤

1. 用docker简化安装。请见主页"安装教程"下的"通过docker安装"章节。910b选择此docker

docker pull swr.cn-central-221.ovaijisuan.com/mindocr/mindocr_dev_ms_2_2_10_cann7_0_py39:v1

2. 进入docker，配置环境变量
3. 拉取mindocr主仓代码
4. cd tools，执行以下命令，自动下载，转换ppocrv4模型。

   bash paddle2mindir.sh -m=ch_PP-OCRv4
   

   详细见链接：https://github.com/mindspore-lab/mindocr/blob/main/docs/cn/inference/inference_thirdparty_quickstart.md 下的"3.5 快速转换工具"章节
5. 执行推理，见 https://github.com/mindspore-lab/mindocr/blob/main/docs/cn/inference/inference_thirdparty_quickstart.md 下的"3.4 端到端推理"章节

[bltcn]

非常感谢，已经测试通过，另外，想请教一下，如果想通过fastapi将其作为服务端接口提供出去，讲模型载入内存后常驻，请问如何做呢，是参考deploy/py_infer/example /ocr_infer_server.py这个嘛？