Toybrick

RK3566,将onnx模型转换成rknn模型时报错

white

新手上路

积分
21
楼主
发表于 2021-9-18 17:59:46    查看: 4079|回复: 0 | [复制链接]    打印 | 只看该作者
环境列表如下:name: rk3566name: '3566'
channels:
  - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
  - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free
  - defaults
dependencies:
  - _libgcc_mutex=0.1=main
  - _openmp_mutex=4.5=1_gnu
  - argon2-cffi=20.1.0=py36h27cfd23_1
  - async_generator=1.10=py36h28b3542_0
  - attrs=21.2.0=pyhd3eb1b0_0
  - bleach=4.0.0=pyhd3eb1b0_0
  - ca-certificates=2021.7.5=h06a4308_1
  - certifi=2021.5.30=py36h06a4308_0
  - cffi=1.14.6=py36h400218f_0
  - dbus=1.13.18=hb2f20db_0
  - defusedxml=0.7.1=pyhd3eb1b0_0
  - entrypoints=0.3=py36_0
  - expat=2.4.1=h2531618_2
  - fontconfig=2.13.1=h6c09931_0
  - freetype=2.10.4=h5ab3b9f_0
  - glib=2.69.1=h5202010_0
  - gst-plugins-base=1.14.0=h8213a91_2
  - gstreamer=1.14.0=h28cd5cc_2
  - icu=58.2=he6710b0_3
  - importlib_metadata=4.6.4=hd3eb1b0_0
  - ipykernel=5.3.4=py36h5ca1d4c_0
  - ipython=6.1.0=py36_0
  - ipython_genutils=0.2.0=pyhd3eb1b0_1
  - ipywidgets=7.6.3=pyhd3eb1b0_1
  - jedi=0.18.0=py36h06a4308_1
  - jinja2=3.0.1=pyhd3eb1b0_0
  - jpeg=9b=h024ee3a_2
  - jsonschema=3.2.0=pyhd3eb1b0_2
  - jupyter=1.0.0=py36_7
  - jupyter_client=7.0.1=pyhd3eb1b0_0
  - jupyter_console=5.2.0=py36_0
  - jupyter_core=4.7.1=py36h06a4308_0
  - jupyterlab_pygments=0.1.2=py_0
  - jupyterlab_widgets=1.0.0=pyhd3eb1b0_1
  - ld_impl_linux-64=2.35.1=h7274673_9
  - libffi=3.3=he6710b0_2
  - libgcc-ng=9.3.0=h5101ec6_17
  - libgomp=9.3.0=h5101ec6_17
  - libpng=1.6.37=hbc83047_0
  - libsodium=1.0.18=h7b6447c_0
  - libstdcxx-ng=9.3.0=hd4cf53a_17
  - libuuid=1.0.3=h1bed415_2
  - libxcb=1.14=h7b6447c_0
  - libxml2=2.9.12=h03d6c58_0
  - markupsafe=2.0.1=py36h27cfd23_0
  - mistune=0.8.4=py36h7b6447c_0
  - nbclient=0.5.3=pyhd3eb1b0_0
  - nbconvert=6.0.7=py36_0
  - nbformat=5.1.3=pyhd3eb1b0_0
  - ncurses=6.2=he6710b0_1
  - nest-asyncio=1.5.1=pyhd3eb1b0_0
  - notebook=6.4.3=py36h06a4308_0
  - openssl=1.1.1l=h7f8727e_0
  - packaging=21.0=pyhd3eb1b0_0
  - pandoc=2.12=h06a4308_0
  - pandocfilters=1.4.3=py36h06a4308_1
  - parso=0.8.2=pyhd3eb1b0_0
  - pcre=8.45=h295c915_0
  - pexpect=4.8.0=pyhd3eb1b0_3
  - pickleshare=0.7.5=pyhd3eb1b0_1003
  - pip=21.0.1=py36h06a4308_0
  - prometheus_client=0.11.0=pyhd3eb1b0_0
  - prompt_toolkit=1.0.15=py36_0
  - ptyprocess=0.7.0=pyhd3eb1b0_2
  - pycparser=2.20=py_2
  - pygments=2.10.0=pyhd3eb1b0_0
  - pyparsing=2.4.7=pyhd3eb1b0_0
  - pyqt=5.9.2=py36h05f1152_2
  - pyrsistent=0.17.3=py36h7b6447c_0
  - python=3.6.13=h12debd9_1
  - python-dateutil=2.8.2=pyhd3eb1b0_0
  - pyzmq=22.2.1=py36h295c915_1
  - qt=5.9.7=h5867ecd_1
  - qtconsole=5.1.0=pyhd3eb1b0_0
  - qtpy=1.10.0=pyhd3eb1b0_0
  - readline=8.1=h27cfd23_0
  - send2trash=1.5.0=pyhd3eb1b0_1
  - setuptools=52.0.0=py36h06a4308_0
  - simplegeneric=0.8.1=py36_2
  - sip=4.19.8=py36hf484d3e_0
  - six=1.16.0=pyhd3eb1b0_0
  - sqlite=3.36.0=hc218d9a_0
  - terminado=0.9.4=py36h06a4308_0
  - testpath=0.5.0=pyhd3eb1b0_0
  - tk=8.6.10=hbc83047_0
  - tornado=6.1=py36h27cfd23_0
  - traitlets=4.3.3=py36_0
  - typing_extensions=3.10.0.0=pyhca03da5_0
  - wcwidth=0.2.5=pyhd3eb1b0_0
  - webencodings=0.5.1=py36_1
  - wheel=0.37.0=pyhd3eb1b0_1
  - widgetsnbextension=3.5.1=py36_0
  - xz=5.2.5=h7b6447c_0
  - zeromq=4.3.4=h2531618_0
  - zipp=3.5.0=pyhd3eb1b0_0
  - zlib=1.2.11=h7b6447c_3
  - pip:
    - absl-py==0.13.0
    - amply==0.1.4
    - astor==0.8.1
    - cached-property==1.5.2
    - chardet==3.0.4
    - cycler==0.10.0
    - dataclasses==0.8
    - decorator==4.4.2
    - docutils==0.17.1
    - flatbuffers==2.0
    - future==0.18.2
    - gast==0.5.2
    - google-pasta==0.2.0
    - grpcio==1.40.0
    - h5py==3.1.0
    - idna==2.8
    - imageio==2.9.0
    - importlib-metadata==4.8.1
    - keras-applications==1.0.8
    - keras-preprocessing==1.1.2
    - kiwisolver==1.3.1
    - markdown==3.3.4
    - matplotlib==3.3.4
    - networkx==2.5.1
    - numpy==1.16.6
    - onnx==1.7.0
    - onnx-tf==1.5.0
    - onnxoptimizer==0.1.0
    - onnxruntime==1.6.0
    - opencv-python==4.4.0.46
    - pillow==8.0.1
    - protobuf==3.12.0
    - psutil==5.6.2
    - pulp==2.4
    - pywavelets==1.1.1
    - pyyaml==5.4.1
    - requests==2.21.0
    - rknn-toolkit2==1.1.0b1
    - ruamel-yaml==0.15.81
    - scikit-image==0.17.2
    - scipy==1.2.1
    - tensorboard==1.14.0
    - tensorflow==1.14.0
    - tensorflow-estimator==1.14.0
    - termcolor==1.1.0
    - tflite==2.3.0
    - tifffile==2020.9.3
    - torch==1.6.0
    - torchvision==0.7.0
    - tqdm==4.27.0
    - typing-extensions==3.10.0.2
    - urllib3==1.24.3
    - werkzeug==2.0.1
    - wrapt==1.12.1
prefix: /home/white/anaconda3/envs/3566

rknn-toolkit版本为rknn-toolkit2==1.1.0b1
我的代码如下:
   onnx_model = './yolox.onnx'
    rknn_model = './yolox.rknn'
    input_size_list = [[1,3,416,416]]
    rknn = RKNN()
    rknn.config(mean_values=[256, 256, 256],std_values=[58.395,58.395,58.395])
    ret = rknn.load_onnx(model=onnx_model)
    ret = rknn.build(do_quantization=True,dataset='./dataset.txt')
    ret = rknn.export_rknn(rknn_model)
    print("down")

输出信息为:
Analysing : 100%|███████████████████████████████████████████████| 268/268 [00:00<00:00, 3452.82it/s]
Quantizating : 100%|█████████████████████████████████████████████| 268/268 [00:02<00:00, 122.62it/s]
I RKNN: librknnc version: 1.1.0b0 (8d7e25ad@2021-06-30T18:33:39)
I RKNN: set log level to 0

Process finished with exit code 136 (interrupted by signal 8: SIGFPE)

并没有得到我预期的结果,并且没有报错信息,只有程序异常终止。请问这是为什么呢?



回复

使用道具 举报

您需要登录后才可以回帖 登录 | 立即注册

本版积分规则

产品中心 购买渠道 开源社区 Wiki教程 资料下载 关于Toybrick


快速回复 返回顶部 返回列表