go-tflite

September 20, 2026 · View on GitHub

Go binding for TensorFlow Lite

Try it in your browser

Click to launch a Jupyter notebook with go-tflite preinstalled and train the iris classifier right in the browser, no install needed. The same image is available as ghcr.io/mattn/go-tflite/gophernotes (see docker/gophernotes).

Usage

model := tflite.NewModelFromFile("sin_model.tflite")
if model == nil {
	log.Fatal("cannot load model")
}
defer model.Delete()

options := tflite.NewInterpreterOptions()
defer options.Delete()

interpreter := tflite.NewInterpreter(model, options)
defer interpreter.Delete()

interpreter.AllocateTensors()

v := float64(1.2) * math.Pi / 180.0
input := interpreter.GetInputTensor(0)
input.Float32s()[0] = float32(v)
interpreter.Invoke()
got := float64(interpreter.GetOutputTensor(0).Float32s()[0])

See _example for more examples

Requirements

TensorFlow Lite C API (libtensorflowlite_c.so). CI builds and tests against TensorFlow v2.17.1; other recent versions should work as long as the C API is compatible.

Tensorflow Installation

go-tflite links against libtensorflowlite_c.so only, so there is no need to build the full TensorFlow library. There are three ways to get it.

Prebuilt buildkit (Linux x86_64)

Each release ships a go-tflite-buildkit-<tag>.tar.gz containing the headers, libtensorflowlite_c.so and the XNNPACK delegate libraries. Extract it into /usr/local and you are done:

$ curl -fSL -o /tmp/buildkit.tar.gz https://github.com/mattn/go-tflite/releases/download/v1.0.7/go-tflite-buildkit-v1.0.7.tar.gz
$ sudo tar xzf /tmp/buildkit.tar.gz -C /usr/local
$ sudo ldconfig

ci/build-buildkit.sh is the script that produces this tarball, so you can run it yourself for other platforms or TensorFlow versions.

Build with bazel

$ cd /source/directory/tensorflow
$ bazel build -c opt //tensorflow/lite/c:tensorflowlite_c

The XNNPACK delegate has no shared library target upstream; see ci/build-buildkit.sh for how to add one.

Build with cmake

$ cmake -S /source/directory/tensorflow/tensorflow/lite/c -B tflite_build
$ cmake --build tflite_build -j

This produces tflite_build/libtensorflowlite_c.so with XNNPACK compiled in. Because there is no separate libtensorflowlite-delegate_xnnpack.so in this case, build programs that use delegates/xnnpack with -tags xnnpack_builtin so that only libtensorflowlite_c is linked.

There is also Makefile.tflite, a plain Makefile that builds libtensorflowlite_c when placed in tensorflow/lite/c; it is not regularly tested.

Environment variables

If the headers and libraries are not installed in a standard location, tell cgo where to find them:

$ export CGO_CFLAGS=-I/source/directory/tensorflow
$ export CGO_LDFLAGS=-L/path/to/tensorflow/libraries
$ export LD_LIBRARY_PATH=/path/to/tensorflow/libraries

Then run go build on some of the examples.

Edge TPU

To be able to compile and use the EdgeTPU delegate, you need to install the libraries from here: https://github.com/google-coral/edgetpu

There is also a deb package here: https://coral.withgoogle.com/docs/accelerator/get-started/#1-install-the-edge-tpu-runtime

The libraries from should be installed in a system wide library path like /usr/local/lib The include files should be installed somewhere that is accesable from your CGO include path

For x86:

cd /tmp && git clone https://github.com/google-coral/edgetpu.git && \
cp edgetpu/libedgetpu/direct/k8/libedgetpu.so.1.0 /usr/local/lib/libedgetpu.so.1.0 && \
ln -rs /usr/local/lib/libedgetpu.so.1.0 /usr/local/lib/libedgetpu.so.1 && \
ln -rs /usr/local/lib/libedgetpu.so.1.0 /usr/local/lib/libedgetpu.so && \
mkdir -p /usr/local/include/libedgetpu && \
cp edgetpu/libedgetpu/edgetpu.h /usr/local/include/edgetpu.h && \
cp edgetpu/libedgetpu/edgetpu_c.h /usr/local/include/edgetpu_c.h && \
rm -Rf edgetpu