TensorFlow Lite is TensorFlow's lightweight solution for Swift developers. It enables low-latency inference of on-device machine learning models with a small binary size and fast performance supporting hardware acceleration.

Build TensorFlow with iOS support

To build the Swift TensorFlow Lite library on Apple platforms, install from source or clone the GitHub repo. Then, configure TensorFlow by navigating to the root directory and executing the configure.py script:

python configure.py

Follow the prompts and when asked to build TensorFlow with iOS support, enter y.

CocoaPods developers

Add the TensorFlow Lite pod to your Podfile:

pod 'TensorFlowLiteSwift'

Then, run pod install.

In your Swift files, import the module:

import TensorFlowLite

Bazel developers

In your BUILD file, add the TensorFlowLite dependency to your target:

swift_library(
  deps = [
      "//tensorflow/lite/swift:TensorFlowLite",
  ],
)

In your Swift files, import the module:

import TensorFlowLite

Build the TensorFlowLite Swift library target:

bazel build tensorflow/lite/swift:TensorFlowLite

Build the Tests target:

bazel test tensorflow/lite/swift:Tests --swiftcopt=-enable-testing

Note: --swiftcopt=-enable-testing is required for optimized builds (-c opt).

Generate the Xcode project using Tulsi

Open the //tensorflow/lite/swift/TensorFlowLite.tulsiproj using the TulsiApp or by running the generate_xcodeproj.sh script from the root tensorflow directory:

generate_xcodeproj.sh --genconfig tensorflow/lite/swift/TensorFlowLite.tulsiproj:TensorFlowLite --outputfolder ~/path/to/generated/TensorFlowLite.xcodeproj