# New Features The following sections describe new features and improvements in the most recent versions of Core ML Tools. ## New in Core ML Tools 8 - Support for [Stateful](stateful-models) Core ML models - Support for [Multifunction](multifunction-models) Core ML models - Several new features in model compression, to see a full list check out the [whats new](opt-whats-new) page in the optimization section. ## Previous Versions The [coremltools 7](https://github.com/apple/coremltools/releases/tag/7.2) package includes APIs for [optimizing the models](opt-overview) to use less storage space, reduce power consumption, and reduce latency during inference. Key optimization techniques include pruning, quantization, and palettization. The [coremltools 6](https://github.com/apple/coremltools/releases/tag/6.3) package offers the following features to optimize the model conversion process: - Model compression utilities; see [Compressing Neural Network Weights](quantization-neural-network). - Float 16 input/output types including image. See [Image Input and Output](image-inputs). For a full list of changes from `coremltools` 5.2, see [Release Notes](#release-notes). ## Release Notes Learn about changes to the `coremltools` package from the following release notes: - [Release Notes (newest)](https://github.com/apple/coremltools/releases/) For information about previous releases, see the following: - [Release Notes for coremltools 7.2](https://github.com/apple/coremltools/releases/tag/7.2) - [Release Notes for coremltools 6.3](https://github.com/apple/coremltools/releases/tag/6.3) - [Release Notes for coremltools 5.2](https://github.com/apple/coremltools/releases/tag/5.2) - [Release Notes for coremltools 5.0](https://github.com/apple/coremltools/releases/tag/5.0) - [Release Notes for coremltools 4.1](https://github.com/apple/coremltools/releases/tag/4.1) - [Release Notes for coremltools 4.0](https://github.com/apple/coremltools/releases/tag/4.0) - [Release Notes for coremltools 3.4](https://github.com/apple/coremltools/releases/tag/3.4) - [All release notes](https://github.com/apple/coremltools/releases) ## Migration Workflow (Core ML Tools 3 → 4) If you used `coremltools` 3 for neural network model conversion from TensorFlow or ONNX/PyTorch to Core ML, update your workflow as follows when you upgrade to `coremltools` 4 and newer: | Conversion from | coremltools 3 | coremltools 4 and newer | | :-------------- | :---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | :----------------------------------------------------------------------------------------------- | | TensorFlow | Install [coremltools 3.4](https://pypi.org/project/coremltools/3.4/) and [tfcoreml 1.1](https://pypi.org/project/tfcoreml/1.1/) and use the `tfcoreml.convert` API. | Use the new `coremltools.convert` API. See [Unified Conversion API](unified-conversion-api). | | PyTorch | First export the PyTorch model to the [ONNX](https://github.com/onnx/onnx) format and then install [coremltools 3.4](https://pypi.org/project/coremltools/3.4/) and [onnx-coreml 1.3](https://pypi.org/project/onnx-coreml/1.3/) and use the `onnx_coreml.convert` API. | Use the new `coremltools.convert` API. See [Unified Conversion API](unified-conversion-api). | ### Convert from TensorFlow With `coremltools` 4 and newer versions, you do not need to install the [tfcoreml](https://github.com/tf-coreml/tf-coreml) package to convert TensorFlow models. The TensorFlow converter is fully integrated in coremltools and available in the [Unified Conversion API](unified-conversion-api). ```{admonition} For older deployment targets To deploy the Core ML model to a target that is iOS 12, macOS 10.13, watchOS 5, tvOS 12, or an older version, use [coremltools 3 and tfcoreml 1](https://github.com/apple/coremltools/blob/3.4/examples/NeuralNetworkGuide.md#tensorflow-conversion). ``` ### Convert from PyTorch You can directly convert from PyTorch using the newest version of coremltools, which includes a PyTorch converter available through the [Unified Conversion API](unified-conversion-api). You no longer need to use the two-step process for converting PyTorch models using the [ONNX](https://github.com/onnx/onnx) format. ```{admonition} For older deployment targets To deploy the Core ML model to a target that is iOS 12, macOS 10.13, watchOS 5, tvOS 12, or an older version, use [coremltools 3 and onnx-coreml 1](https://github.com/apple/coremltools/blob/3.4/examples/NeuralNetworkGuide.md#onnx-converter) ``` ### Deprecated Methods and Support In coremltools 4 and newer, the following class and methods available in previous versions are deprecated: - `convert_neural_network_weights_to_fp16()`, `convert_neural_network_spec_weights_to_fp16()`, and `quantize_spec_weights()`. Use the `quantize_weights()` method instead. For instructions, see [Quantization](quantization-neural-network). - The NeuralNetworkShaper class. - `get_allowed_shape_ranges()`. - `can_allow_multiple_input_shapes()`. - `visualize_spec()` method of the MLModel class. You can use the [netron](https://github.com/lutzroeder/netron) open source viewer to visualize Core ML models. - `get_custom_layer_names()`, `replace_custom_layer_name()`, and `has_custom_layer()`: These were moved to internal methods. - Caffe converter - Keras.io and ONNX converters will be deprecated in coremltools 6. Users are recommended to transition to the TensorFlow/PyTorch conversion using the [Unified Conversion API](unified-conversion-api). The current version of coremltools ([version 7.1](https://github.com/apple/coremltools)) includes wheels for Python 3.7, 3.8, 3.9, 3.10, and 3.11.