Installation

Installation from source

You can download oneml source code from Gitlab release page , Alternatively cloning the repository using git

git clone git@gitlab.com:sertiscorp/mle/edge/oneml.git

Install all the dependencies locally or use Docker

Building with build script

scripts/build.sh \
    [-f config file] \
    [-t target_arch] \
    [-b build_type] \
    [-bt build_test] \
    [-lc build_with_license] \
    [-ba build_sample_apps] \
    [-bd build_documentation] \
    [-py python_bindings] \
    [-jni java_bindings] \
    [-go go_bindings] \
    [--clean clean_build] \

To see all available options ./scripts/build.sh --help

Supported model format

• MXnet

• Pytorch

• CAFFE

• Tensorflow

• ONNX

• Darknet

• MegEngine

Note

Model compatibility is based on model format support by the framework. Not all models can be converted, compatibility needed to be verified manually.

Docker

Dockerfile containing the dependencies is located at docker/Dockerfile.

Docker image with all dependencies can be found at container registry or simply run

For CPU image
docker pull registry.gitlab.com/sertiscorp/mle/edge/oneml:develop

For GPU image
docker pull registry.gitlab.com/sertiscorp/mle/edge/oneml:develop-gpu

For Android image
docker pull registry.gitlab.com/sertiscorp/mle/edge/oneml:develop-android