TensorRT7 OSS on JetPack 4.4 Developer Preview

Hi everyone,

I am using the Jetpack 4.4 developer preview with
TensorRT 7.1.0, cuda 10.2, cmake v3.16.6

I want to build and install the TensorRT 7.0 OSS on the Jetson AGX Xavier Developer kit.
I am following these instructions from Mask R-CNN network in TensorRT:

  1. sudo git clone https://github.com/nvidia/TensorRT TensorRT -b release/7.0
  2. cd TensorRT
  3. sudo git submodule update --init --recursive
  4. sudo mkdir -p build && cd build
  5. sudo cmake -DCUBLASLT_LIB=/usr/local/cuda-10.2/targets/aarch64-linux/lib/libcublas.so \
    -DCMAKE_CUDA_COMPILER=/usr/local/cuda-10.2/bin/nvcc \
    -DGPU_ARCHS=72 ..
  6. sudo make

It is not able to build the “libnvinfer_plugin” and it gives the following messages:
[ 18%] Building CUDA object plugin/CMakeFiles/nvinfer_plugin.dir/common/kernels/proposalsForward.cu.o
[ 18%] Building CUDA object plugin/CMakeFiles/nvinfer_plugin.dir/common/kernels/regionForward.cu.o
[ 20%] Building CUDA object plugin/CMakeFiles/nvinfer_plugin.dir/common/kernels/reorgForward.cu.o
[ 20%] Building CUDA object plugin/CMakeFiles/nvinfer_plugin.dir/common/kernels/roiPooling.cu.o
[ 20%] Building CUDA object plugin/CMakeFiles/nvinfer_plugin.dir/common/kernels/rproiInferenceFused.cu.o
[ 21%] Building CUDA object plugin/CMakeFiles/nvinfer_plugin.dir/common/kernels/sortScoresPerClass.cu.o
[ 21%] Building CUDA object plugin/CMakeFiles/nvinfer_plugin.dir/common/kernels/sortScoresPerImage.cu.o
[ 21%] Building CUDA object plugin/CMakeFiles/nvinfer_plugin.dir/common/cudaDriverWrapper.cu.o
[ 21%] Building CUDA object plugin/CMakeFiles/nvinfer_plugin.dir/embLayerNormPlugin/embLayerNormPlugin.cu.o
[ 22%] Building CUDA object plugin/CMakeFiles/nvinfer_plugin.dir/fcPlugin/fcPlugin.cu.o
[ 22%] Building CUDA object plugin/CMakeFiles/nvinfer_plugin.dir/geluPlugin/geluPlugin.cu.o
[ 22%] Building CUDA object plugin/CMakeFiles/nvinfer_plugin.dir/bertQKVToContextPlugin/qkvToContextPlugin.cu.o
[ 23%] Building CUDA object plugin/CMakeFiles/nvinfer_plugin.dir/skipLayerNormPlugin/skipLayerNormPlugin.cu.o
[ 23%] Building CXX object plugin/CMakeFiles/nvinfer_plugin.dir/InferPlugin.cpp.o
[ 23%] Building CXX object plugin/CMakeFiles/nvinfer_plugin.dir/__/samples/common/logger.cpp.o
make[2]: *** No rule to make target ‘/usr/local/cuda-10.2/targets/aarch64-linux/lib/libcublas.so’, needed by ‘libnvinfer_plugin.so.7.0.0.1’. Stop.
CMakeFiles/Makefile2:1333: recipe for target ‘plugin/CMakeFiles/nvinfer_plugin.dir/all’ failed
make[1]: *** [plugin/CMakeFiles/nvinfer_plugin.dir/all] Error 2
Makefile:154: recipe for target ‘all’ failed
make: *** [all] Error 2

Any help would be greatly appreciated

I was able to build TensorRT 7.0 OSS using Build Jetson TensorRT OSS Plugin
My problem was related to the cmake’s parameters:

sudo /usr/local/bin/cmake .. \
-DGPU_ARCHS=72 \
-DCMAKE_CUDA_COMPILER=/usr/local/cuda-10.2/bin/nvcc \
-DTRT_LIB_DIR=/usr/lib/aarch64-linux-gnu/ \
-DCMAKE_C_COMPILER=/usr/bin/gcc \
-DTRT_BIN_DIR=pwd/out