Skip to content

NNEF-help.md

Latest commit

 

History

History
47 lines (41 loc) · 2.12 KB

NNEF-help.md

File metadata and controls

47 lines (41 loc) · 2.12 KB
Raw

Neural Network Exchange Format (NNEF)

NNEF reduces machine learning deployment fragmentation by enabling a rich mix of neural network training tools and inference engines to be used by applications across a diverse range of devices and platforms.

Steps to setup NNEF-Tools on RPi

  • Note: Python3 version >= 3.5.2 is supported

Step 1: Download and install NNEF Tools:

git clone https://github.com/KhronosGroup/NNEF-Tools
cd NNEF-Tools
cd parser/python
python setup.py install
cd ../..

Step 2: Install dependencies

pip install future typing six numpy protobuf onnx onnx-simplifier onnxruntime

Step 3: Convert model from ONNX to NNEF

python -m nnef_tools.convert --help
usage: convert.py [-h] --input-model INPUT_MODEL [--output-model OUTPUT_MODEL]
                  --input-format {tf,tflite,onnx,nnef,caffe2,caffe}
                  --output-format {tf,tflite,onnx,nnef,caffe2}
                  [--input-shapes INPUT_SHAPES]
                  [--io-transpose [IO_TRANSPOSE [IO_TRANSPOSE ...]]]
                  [--fold-constants] [--optimize]
                  [--custom-converters CUSTOM_CONVERTERS [CUSTOM_CONVERTERS ...]]
                  [--custom-shapes CUSTOM_SHAPES [CUSTOM_SHAPES ...]]
                  [--custom-fragments CUSTOM_FRAGMENTS [CUSTOM_FRAGMENTS ...]]
                  [--custom-optimizers CUSTOM_OPTIMIZERS [CUSTOM_OPTIMIZERS ...]]
                  [--mirror-unsupported] [--generate-custom-fragments]
                  [--keep-io-names] [--decompose [DECOMPOSE [DECOMPOSE ...]]]
                  [--input-names INPUT_NAMES [INPUT_NAMES ...]]
                  [--output-names OUTPUT_NAMES [OUTPUT_NAMES ...]]
                  [--tensor-mapping [TENSOR_MAPPING]] [--annotate-shapes]
                  [--compress [COMPRESS]]
  • Command:
python -m nnef_tools.convert --input-format onnx --output-format nnef --input-model <PATH-TO-MNIST-ONNX-MODEL>/mnist.onnx --output-model <PATH-TO-SAVE-NNEF-MODEL>/mnist.nnef