ServiceNow completed its acquisition of Element AI on January 8, 2021. All references to Element AI in the materials that are part of this project should refer to ServiceNow.

build-pynini-wheels

Build manylinux2014_x86_64 Python wheels for pynini, wrapping all its dependencies.

This project heavily relies on other open-source projects.

• See https://github.com/pypa/manylinux/tree/manylinux2014 for information about manylinux2014.
• See http://www.opengrm.org/twiki/bin/view/GRM/Pynini for information about pynini.
• See http://www.openfst.org/twiki/bin/view/FST/WebHome for information about OpenFst.

What is this?

As of the writing of these lines, the recommended installation method for Pynini was through Conda-Forge. The enclosed Dockerfile gives you another alternative: build your own Python Platform Wheels for pynini so that you may easily pip/poetry-install it in your favourite linux.

The process differs for macOS and Windows.

Usage

To build wheels and run Pynini's tests for x86_64, run:

docker build \
    --build-arg="PYNINI_VERSION=2.1.5" \
    --build_arg="FST_VERSION=1.8.2" \
    --target=run-tests \
    -t build-pynini-wheels \
    .

To build wheels and run Pynini's tests for aarch64, run:

docker build \
    --build-arg="PYNINI_VERSION=2.1.5" \
    --build_arg="FST_VERSION=1.8.2" \
    --target=run-tests \
    -t build-pynini-wheels \
    -f Dockerfile.arm64 \
    .

To extract the resulting wheels from the Docker image, run:

docker run --rm -v `pwd`:/io build-pynini-wheels cp -r /wheelhouse /io

Notice that this may also give you Cython wheels.

To publish these wheels to a PyPI repository (either a local one or the official https://pypi.org/ ), you may be interested in twine (see https://twine.readthedocs.io/en/latest/ ).

Maintenance?

This repository aims to create its own obsolescence by providing the pynini maintainer with the means to create their own wheels. Updates may thus be "spotty" at best, and the repository will be retired once an official alternative is available.