How to convert MXNet model to Apple CoreML:
docker run -v "$PWD":/data --rm -it dharmeshkakadia/mxnet-coreml-tools-docker python mxnet_coreml_converter.py
For example, if you want to convert Squeezenet model to coreml, to use with iOS.
Run the following from the a directory containing Squeezenet model files (Params, symbols, labels) and will generate squeezenetv11.mlmodel in the current directory.
docker run -v "$PWD":/data --rm -it dharmeshkakadia/mxnet-coreml-tools-docker python mxnet_coreml_converter.py --model-prefix='/data/squeezenet_v1.1' --epoch=0 --input-shape='{"data":"3,227,227"}' --mode=classifier --pre-processing-arguments='{"image_input_names":"data"}' --class-labels /data/synset.txt --output-file="/data/squeezenetv11.mlmodel"
The Dockerfile is located on github.
