Running CUDA, CuDNN, TensorRT, or Driveworks Sample Apps on Target

  1. Run CUDA applications:

    CUDA samples are packaged at /home/root/samples/cuda.

    ./sample-application>
  2. Run CuDNN applications:

    CuDNN samples are packaged at /home/root/samples/cudnn 

    ./<sample-application>
  3. Run TensorRT applications:

    TensorRT samples are packaged at /home/root/samples/tensorrt

    The following sample applications are available:

    sample_char_rnn
sample_googlenet
sample_movielens
sample_fasterRCNN

    sample_fasterRCNN requires the TensorRT model file to be present in the data/faster-rcnn/ subfolder. Download the model file here:

    https://dl.dropboxusercontent.com/s/o6ii098bu51d139/faster_rcnn_models.tgz?dl=0

  4. Run the NVIDIA DriveWorks applications:

    The NVIDIA DriveWorks 'core' samples are packaged at /usr/local/driveworks-<DW-VERSION>/bin/sample.

    ./<sample-application>

    If DriveWorks samples fail, export LD_LIBRARY_PATH:

    export LD_LIBRARY_PATH=/usr/local/driveworks-<DW-VERSION>/lib:/usr/local/cuda-<CUDA-VERSION>/targets/aarch64-linux/lib
    Note: The test data files required by some of the DriveWorks samples are not packaged in the rootfs by default, due to their large size. Samples from DW add-on packages are also not packaged.

    In order to run the DriveWorks sample apps that require test data files, the scp utility may be used to push the required files into the target at /usr/local/driveworks-<DW-VERSION>/data/.

    The additional DriveWorks samples may be pushed via scp onto the target, at /usr/local/driveworks-<DW-VERSION>/bin/sample/.