Core ML Community Tools
Core ML community tools contains all supporting tools for Core ML model
conversion, editing and validation. This includes deep learning frameworks like
TensorFlow, Keras, Caffe as well as classical machine learning frameworks like
LIBSVB, scikit-learn, and XGBoost.
To get the latest version of coremltools:
pip install --upgrade coremltools
For the latest changes please see the release notes.
Table of Contents
- Neural network conversion
- Core ML specification
- coremltools user guide and examples
- Installation from Source
Neural Network Conversion
Link to the detailed NN conversion guide.
There are several converters available to translate neural networks trained
in various frameworks into the Core ML model format. Following formats can be
converted to the Core ML .mlmodel format through the coremltools python
package (this repo):
- Caffe V1 (
.prototxt,.caffemodelformat) - Keras API (2.2+) (
.h5format) - TensorFlow 1 (1.13+) (
.pbfrozen graph def format) - TensorFlow 2 (
.h5andSavedModelformats)
In addition, there are two more neural network converters build on top of coremltools:
- onnx-coreml: to convert
.onnxmodel format. Several frameworks such as PyTorch, MXNet, CaffeV2 etc
provide native export to the ONNX format. - tfcoreml: to convert TensorFlow models. For producing Core ML models targeting iOS 13 or later,
tfcoreml defers to the TensorFlow converter implemented inside coremltools.
For iOS 12 or earlier, the code path is different and lives entirely in the tfcoreml package.
To get an overview on how to use the converters and features such as
post-training quantization using coremltools, please see the neural network
guide.
Core ML Specification
- Core ML specification is fully described in a set of protobuf files.
They are all located in the foldermlmodel/format/ - For an overview of the Core ML framework API, see here.
- To find the list of model types supported by Core ML, see this
portion of themodel.protofile. - To find the list of neural network layer types supported see this
portion of theNeuralNetwork.protofile. - Auto-generated documentation for all the protobuf files can be found at this link
User Guide and Examples
- API documentation
- Updatable models
- Neural network inference examples
- Neural network guide
- Miscellaneous How-to code snippets
Installation
We recommend using virtualenv to use, install, or build coremltools. Be
sure to install virtualenv using your system pip.
pip install virtualenv
The method for installing coremltools follows the
standard python package installation steps.
To create a Python virtual environment called pythonenv follow these steps:
# Create a folder for your virtualenv
mkdir mlvirtualenv
cd mlvirtualenv
# Create a Python virtual environment for your Core ML project
virtualenv pythonenv
To activate your new virtual environment and install coremltools in this
environment, follow these steps:
# Active your virtual environment
source pythonenv/bin/activate
# Install coremltools in the new virtual environment, pythonenv
(pythonenv) pip install -U coremltools
The package documentation contains
more details on how to use coremltools.