Converts a PaddlePaddle model (ProgramDesc + parameters) into an ONNX graph. Uses the ONNX pip library and targets PaddlePaddle Fluid. Built in Python 2.7 (and underneath the hood, ONNX does a Pybind to their C++ libraries).
To understand PaddlePaddle's (non-)graph way of representing a deep learning program, a ProgramDesc, refer to: https://github.com/PaddlePaddle/Paddle/blob/develop/doc/fluid/design/concepts/program.md.
Targets Paddle->ONNX conversion for now, and will consequently support the reverse too.
Currently a work-in-progress tool since there a features in PaddlePaddle not supported in ONNX today and vice-versa.
First, generate model directory by running any fluid test / example and write the model using the fluid.io.save_inference_model API.
Then, run convert.py by providing the generated model directory to the argument ---modeldir.
(TBD)
If you don't already have protobuf installed on your computer, install it from here: https://github.com/google/protobuf. On Mac, to get the development version, use brew install protobuf.
Create a virtual environment and install ONNX using PIP.
virtualenv venv
source venv/bin/activate
pip install -r requirements.txt
Build PaddlePaddle's develop branch from source using info here:
http://paddlepaddle.org/docs/develop/documentation/en/build_and_install/build_from_source_en.html. Make the paddle/python available in the execution environment's PYTHONPATH, or pip install the wheel after building the target paddle_python.
NOTE: Make sure your virtual environment has the new Protobuf used by this project and the onnx dependency, as Paddle installation may try to downgrade it.
TBD
We aim to at least support all the models from our model bank. During our preliminary stage, we plan to support the models generated from:
Provided under the Apache-2.0 license.