Releases: tf-coreml/tf-coreml
tfcoreml 2.0.0
Changes
-
minimum_ios_deployment_target
can now only be one of['12', '11.2']
. -
Deprecated conversion support for TensorFlow 2.
tfcoreml
now only accepts TensorFlow 1 models serialized as frozengraph_def
object.
For TensorFlow 2 conversion or when minimum deployment target is iOS13
or later, please use Unified Conversion API introduced in the coremltools 4.0
python package.
tfcoreml 1.1.0
Changes:
- Add support for TensorFlow 2.x file format (.h5, SavedModel, and concrete functions).
- Add support for several new ops, such as
AddV2
,FusedBatchNormV3
for iOS 13+. - Bug fixes in the TensorFlow converter's op fusion graph pass.
- New
minimum_ios_deployment_target
flag (replacing previoustarget_ios
which causes confusions). This flag can be utilized for setting minimum targeting iOS, e.g., the model converted withminimum_ios_deployment_target='12'
will support iOS 12, iOS 13, and later.
Known Issues:
tf.keras
model conversion supported only with TensorFlow 2- Currently, there are issues while invoking the TensorFlow 2.x model conversion in Python 2.x.
- Currently, there are issues while converting
tf.keras
graphs that contain recurrent layers.
tfcoreml 1.0.0
This release was focused on adding many new layers and operators which can be found here. The coremltools release 3.0 can be found here.
Examples for model conversions and how to use control flows
- Added the argument
target_ios
to choose the Core ML spec version that is produced by the converter.target_ios='13'
will enable the converter to use all the new layers added in Core ML 3. - Added
custom_conversion_function
option where users can provide their own custom conversion function similar to these examples for the custom conversion use case.
Supports new model specification version 4 with target iOS 13 or later, the Core ML model specification can be found in proto or documentation.
Any questions or concerns related to this release can be submitted as an issue and will be review by the team. All comments are welcomed and will be used to improve the existing documentation.
tfcoreml 0.3.0
- added support for resizeBilinear and cropAndResize ops
- added support for 8-bit QuantizedConv2D, QuantizedMatMul ops
- some bug fixes
tfcoreml 0.2.0
- Adds Python 3 support
- Adds custom layer support
- Miscellaneous bug fixes