Unofficial TensorFlow cocoapod for iOS with support for inference, evaluation, and training. Pod targets simluator and arm64 devices only (iOS 12.0+) with full support for training MobileNetV2 models on device. We use this pod in TensorIO.
- Latest: v2.0.9
- TensorFlow Build: r2.0, v2.0.3
The major.minor version number of this pod tracks the major.minor version of the build of TensorFlow it includes. We reserve our patch numbers for our own changes to the build, which normally involves whitelisting additinal ops to support new models.
- Xcode 10.0+
- iOS 12.0+
TensorIOTensorFlow is available through CocoaPods. To install it, simply add the following line to your Podfile:
pod 'TensorIOTensorFlow'
As of v2.0.6 the pod vends only x86_64 and arm64 builds of tensorflow. You may need to add arm4
to your application's Required device capabilities in the Target > Info > Custom iOS Target Properties tab of your project. You may also need to update the Build Settings > Architectures for the project to arm64
.
Based on doc.ai's tensorflow-ios-framework, this pod vends all the headers required to use TensorFlow in your project as well as the following static libraries:
For projects that use this pod be sure to add the following to your header search paths and linker flags:
Header Search Paths:
${PODS_ROOT}/TensorIOTensorFlow/Frameworks/tensorflow.framework/Headers
Linker Flags:
-force_load "${PODS_ROOT}/TensorIOTensorFlow/Frameworks/tensorflow.framework/tensorflow"
If you are using Tensor/IO with this backend, these are already added to your project. If you are developing your own pod with TensorIOTensorFlow as a dependency, you can add the following to your podspec to do this automatically:
s.xcconfig = {
'HEADER_SEARCH_PATHS' => '"${PODS_ROOT}/TensorIOTensorFlow/Frameworks/tensorflow.framework/Headers"',
'OTHER_LDFLAGS' => '-force_load "${PODS_ROOT}/TensorIOTensorFlow/Frameworks/tensorflow.framework/tensorflow"'
}
Build this pod from scratch with the build.sh script in this repository.
doc.ai
TensorIOTensorFlow is available under the Apache 2.0 license. See the LICENSE file for more info.