onnx-connx-neowine 0.1.2

ONNX to CONNX converting tool

ONNX to CONNX Converter

onnx-connx is a tool which converts ONNX to CONNX model. And onnx-connx is also a NumPy ONNX Runtime implementation for CONNX. We don't recommend to use onnx-connx as an ONNX Runtime for NumPy is very (very) slow.

For users

Install onnx-connx via pip

pip install git+https://github.com/semihlab/onnx-connx

Convert ONNX to CONNX model

python -m onnx_connx --help        # to get help message
python -m onnx_connx [onnx model]  # to convert onnx to connx

For developers

Prepare development environments

• python3
• onnx # python package, to run onnx2connx converter
• protobuf-compiler # to run bin/dump utility
• tabulate # python package, to run test cases
• pytest-parallel # python package, to run test cases

$ sudo apt install python3 python3-pip python3-venv
$ python3 -m pip install --user virtualenv
$ python3 -m venv venv
$ source venv/bin/activate
$ pip install --upgrade pip
$ pip install onnx

Debug installation

pip install git+file:///[path-to-onnx-connx]

Dump onnx to text

onnx-connx$ bin/dump [onnx path]  # This utility will dump onnx or pb to text using protoc

Test

Run test

onnx-connx$ make test

Convert all test cases that connx supports

onnx-connx$ make convert-test [ONNX_HOME=[onnx install dir]] [CONNX_HOME=[connx source dir]] # ONNX_HOME and CONNX_HOME can be omitted

Convert individual test case

onnx-connx$ bin/convert [onnx test case path] [connx test case path]

Run connx backend

connx backend will compile the ONNX to CONNX and run it using NumPy operators.

python -m onnx_connx.backend [onnx model] [[input tensor] ...]

Run MNIST example

onnx-connx$ cd examples
onnx-connx/examples$ ./download.sh
onnx-connx/examples$ cd ..
onnx-connx$ python -m onnx_connx.backend examples/mnist/model.onnx examples/mnist/input_0.pb

Contribution

See CONTRIBUTING.md

License

CONNX is licensed under GPLv3. See LICENSE