按教程部署完insightface后发现只能使用cpu

使用文档

首先是根据这个文档InsightFace Python部署示例,根据其环境要求导航到FastDeploy RKNPU2 导航文档,

我的环境

python是3.8
rknn版本是2.3.0

使用情况

按照教程指令编译成功

git clone https://github.com/PaddlePaddle/FastDeploy.git
cd FastDeploy

# 如果您使用的是develop分支输入以下命令
git checkout develop

cd python
export ENABLE_ORT_BACKEND=ON
export ENABLE_RKNPU2_BACKEND=ON
export ENABLE_VISION=ON

# 请根据你的开发版的不同，选择RK3588和RK356X
export RKNN2_TARGET_SOC=RK3588

# 如果你的核心板的运行内存大于等于8G，我们建议您执行以下命令进行编译。
python3 setup.py build
# 值得注意的是，如果你的核心板的运行内存小于8G，我们建议您执行以下命令进行编译。
python3 setup.py build -j1

python3 setup.py bdist_wheel
cd dist
pip3 install fastdeploy_python-0.0.0-cp39-cp39-linux_aarch64.whl

部分编译参数如下

安装fastdeploy_python后按照文档指令

#下载部署示例代码
git clone https://github.com/PaddlePaddle/FastDeploy.git
cd examples/vision/faceid/insightface/python/

#下载ArcFace模型文件和测试图片
wget https://bj.bcebos.com/paddlehub/fastdeploy/ms1mv3_arcface_r100.onnx
wget https://bj.bcebos.com/paddlehub/fastdeploy/rknpu2/face_demo.zip
unzip face_demo.zip

# CPU推理
python infer_arcface.py --model ms1mv3_arcface_r100.onnx \
                        --face face_0.jpg \
                        --face_positive face_1.jpg \
                        --face_negative face_2.jpg \
                        --device cpu
# GPU推理
python infer_arcface.py --model ms1mv3_arcface_r100.onnx \
                        --face face_0.jpg \
                        --face_positive face_1.jpg \
                        --face_negative face_2.jpg \
                        --device gpu

gpu版本会提示

/FastDeploy/examples/vision/faceid/insightface/python$ python infer_arcface.py --model ms1mv3_arcface_r100.onnx                         --face face_0.jpg                         --face_positive face_1.jpg                         --face_negative face_2.jpg                         --device gpu
WARNING:root:The installed fastdeploy-python package is not built with GPU, will force to use CPU. To use GPU, following the commands to install fastdeploy-gpu-python.
WARNING:root:    ================= Install GPU FastDeploy===============
WARNING:root:    python -m pip uninstall fastdeploy-python
WARNING:root:    python -m pip install fastdeploy-gpu-python -f https://www.paddlepaddle.org.cn/whl/fastdeploy.html
[INFO] fastdeploy/runtime/runtime.cc(326)::CreateOrtBackend	Runtime initialized with Backend::ORT in Device::CPU.
FaceRecognitionResult: [Dim(512), Min(-2.309219), Max(2.372197), Mean(0.016987)]
FaceRecognitionResult: [Dim(512), Min(-2.288257), Max(1.995103), Mean(-0.003400)]
FaceRecognitionResult: [Dim(512), Min(-3.243412), Max(3.875865), Mean(-0.030682)]
Cosine 01:  0.814384554852488
Cosine 02:  -0.059388045136689285
RuntimeOption(
  backend : Backend.ORT	
  cpu_thread_num : -1	
  device : Device.CPU	
  device_id : 0	
  external_stream : None	
  model_file : ms1mv3_arcface_r100.onnx	
  model_format : ModelFormat.ONNX	
  model_from_memory : False	
  openvino_option : <fastdeploy.libs.fastdeploy_main.OpenVINOBackendOption object at 0x7fa003bc30>	
  ort_option : <fastdeploy.libs.fastdeploy_main.OrtBackendOption object at 0x7fa0078530>	
  paddle_infer_option : <fastdeploy.libs.fastdeploy_main.PaddleBackendOption object at 0x7fa0078530>	
  paddle_lite_option : <fastdeploy.libs.fastdeploy_main.LiteBackendOption object at 0x7fa0078530>	
  params_file : 	
  poros_option : <fastdeploy.libs.fastdeploy_main.PorosBackendOption object at 0x7fa0078530>	
  trt_option : <fastdeploy.libs.fastdeploy_main.TrtBackendOption object at 0x7fa0078530>	
)

输入指令查看npu使用情况确实没变化

Start with examples/vision/faceid/insightface/rknpu2/python/README_CN.md and docs/cn/build_and_install/rknpu2.md, then inspect python/setup.py and examples/vision/faceid/insightface/python/infer_arcface.py. Reproduce the --device gpu invocation and compare it with the reported build configuration; done means the documented RKNPU2 path uses the NPU or the documentation clearly explains the required configuration.