Build Tensorflow Lite shared library:
Modify the code: tensorflow/contrib/lite/BUILD in TensorFlow source:
Build it:
P.S1: From TensorFlow 1.13, the "lite" directory is not under the "tensorflow/contrib" directory, instead directly in "tensorflow/".
P.S2:
you may need to download your TensorFlow version's flatbuffers in here:
https://github.com/tensorflow/tensorflow/blob/master/third_party/flatbuffers/workspace.bzl
"http://mirror.tensorflow.org/github.com/google/flatbuffers/archive/v1.11.0.tar.gz"
or
"https://github.com/google/flatbuffers/archive/v1.11.0.tar.gz",
And build it using CMake:
Modify the code: tensorflow/contrib/lite/BUILD in TensorFlow source:
cc_binary(
name = "libtflite.so",
deps = [":framework",
"//tensorflow/contrib/lite/kernels:builtin_ops",
"//tensorflow/contrib/lite/kernels:eigen_support",
"//tensorflow/contrib/lite/kernels:gemm_support",
],
linkopts=["-shared -Wl,--whole-archive" ],
linkshared=1
)
Build it:
bazel build --config opt //tensorflow/contrib/lite:libtflite.so --cxxopt="-std=c++11" --copt="-O3"
P.S1: From TensorFlow 1.13, the "lite" directory is not under the "tensorflow/contrib" directory, instead directly in "tensorflow/".
P.S2:
you may need to download your TensorFlow version's flatbuffers in here:
https://github.com/tensorflow/tensorflow/blob/master/third_party/flatbuffers/workspace.bzl
"http://mirror.tensorflow.org/github.com/google/flatbuffers/archive/v1.11.0.tar.gz"
or
"https://github.com/google/flatbuffers/archive/v1.11.0.tar.gz",
And build it using CMake:
cmake -G "Unix Makefiles"
sudo make install
No comments:
Post a Comment