tensorflow r1.9 + tensorRT compilation with Jetpack 4.2

I am trying to compile tensorflow r1.9 and tensorRT with bazel 0.19.0. This compilation is for c++ application to be running on Xavier.

I want to use ConvertGraphDefToTensorRT, to convert a frozen tf graph to tensorRT format.

The libarary, bazel-bin/tensorflow/libtensorflow_cc.so which i get after compilation does not incluce function ’ ConvertGraphDefToTensorRT’

I am checking by using below command.
readelf -Ws bazel-bin/tensorflow/libtensorflow_cc.so | grep -r -n ‘ConvertGraphDefToTensorRT’

Below is the content of .tf_configure.bazelrc :
build --action_env PYTHON_BIN_PATH="/usr/bin/python"
build --action_env PYTHON_LIB_PATH="/usr/local/lib/python2.7/dist-packages"
build --python_path="/usr/bin/python"
build --define with_jemalloc=true
build:gcp --define with_gcp_support=true
build:hdfs --define with_hdfs_support=true
build:s3 --define with_s3_support=true
build:kafka --define with_kafka_support=true
build --define with_xla_support=true
build:gdr --define with_gdr_support=true
build:verbs --define with_verbs_support=true
build --action_env TF_NEED_OPENCL_SYCL=“0”
build --action_env TF_NEED_CUDA=“1”
build --action_env CUDA_TOOLKIT_PATH="/usr/local/cuda"
build --action_env TF_CUDA_VERSION=“10.0”
build --action_env CUDNN_INSTALL_PATH="/usr/lib/aarch64-linux-gnu"
build --action_env TF_CUDNN_VERSION=“7”
build --action_env TENSORRT_INSTALL_PATH="/usr/lib/aarch64-linux-gnu"
build --action_env TF_TENSORRT_VERSION=“5.0.3”
build --action_env TF_NCCL_VERSION=“1”
build --action_env TF_CUDA_COMPUTE_CAPABILITIES=“7.0,7.2”
build --action_env TF_CUDA_CLANG=“0”
build --action_env GCC_HOST_COMPILER_PATH="/usr/bin/gcc"
build --config=cuda
test --config=cuda
build --define grpc_no_ares=true
build:opt --copt=-march=native
build:opt --host_copt=-march=native
build:opt --define with_default_optimizations=true
build --strip=always

Hi,

Would you mind to check the TensorRT installation first?
https://docs.nvidia.com/deeplearning/sdk/tensorrt-install-guide/index.html#installing-debian

$ dpkg -l | grep TensorRT

Thanks

TensorRT is installed and libnvinfer5 is there.

Is ConvertGraphDefToTensorRT in libnvinfer libs?

Hi,

After checking, the ConvertGraphDefToTensorRT function is in the TensorFlow:
https://github.com/tensorflow/tensorflow/blob/r1.13/tensorflow/contrib/tensorrt/convert/convert_graph.cc

TensorFlow for Xavier can be installed with the command here:
https://devtalk.nvidia.com/default/topic/1042125/jetson-agx-xavier/official-tensorflow-for-jetson-agx-xavier/

Thanks.