Establishing TensorSpace-Converter development environment is a tedious topic? Dockerize it!

Here is a TensorSpace-Converter Dockerfile, you can use it to build tensorspacejs image, which is an out-of-box TensorSpace-Converter environment. We also provide some easy to use scripts to init (init_docker_converter.sh) and run (run_docker_converter.sh) tensorspacejs docker image. This demo provide a practical usage of TensorSpace-Converter in Docker.

To init tensorspacejs Docker image (make sure start Docker daemon before init the image):

bash init_docker_converter.sh

To run docker image. Put TensorSpace-Converter script and model assets in a work_dir, and execute run_docker_converter.sh to run tensorspacejs image:

bash run_docker_converter.sh --work_dir ./PATH/TO/WORK/DIRECTORY

work_dir is where to place converter.sh and input model assets, meanwhile, tensorspacejs Docker image will save generated model files into this directory. Configure work_dir when run the tensorspacejs Docker image through run_docker_converter.sh. In this example, configure work_dir to be ./example

converter.sh contains TensorSpace-Converter conversion code, tensorspacejs image will execute this script. Place converter.sh at the root of work_dir, make sure the file name is converter.sh (should not be renamed).

Place input model under work_dir, the path of input model and output folder are relative to work_dir. In this example, set input_path to be ./input/keras_model.h5 and output_path to be ./output in converter.sh.