Error messages related to NVIDIA libraries.

[wangkuiyi]

Following https://github.com/sql-machine-learning/elasticdl/blob/develop/docs/tutorials/elasticdl_local.md, I ran the following command to generate datasets in ReocrdIO format:

docker run --rm -it \
  -v $HOME/.keras:/root/.keras \
  -v $PWD:/work \
  -w /work \
  elasticdl/elasticdl:dev bash -c "/scripts/gen_dataset.sh data"

The data generation works. However, it prints some weird error messages.

Status: Downloaded newer image for elasticdl/elasticdl:dev
2020-07-13 19:17:18.151377: W tensorflow/stream_executor/platform/default/dso_loader.cc:55] Could not load dynamic library 'libnvinfer.so.6'; dlerror: libnvinfer.so.6: cannot open shared object file: No such file or directory
2020-07-13 19:17:18.151837: W tensorflow/stream_executor/platform/default/dso_loader.cc:55] Could not load dynamic library 'libnvinfer_plugin.so.6'; dlerror: libnvinfer_plugin.so.6: cannot open shared object file: No such file or directory
2020-07-13 19:17:18.151914: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:30] Cannot dlopen some TensorRT libraries. If you would like to use Nvidia GPU with TensorRT, please make sure the missing libraries mentioned above are installed properly.
2020-07-13 19:17:41.865367: W tensorflow/stream_executor/platform/default/dso_loader.cc:55] Could not load dynamic library 'libnvinfer.so.6'; dlerror: libnvinfer.so.6: cannot open shared object file: No such file or directory
2020-07-13 19:17:41.866023: W tensorflow/stream_executor/platform/default/dso_loader.cc:55] Could not load dynamic library 'libnvinfer_plugin.so.6'; dlerror: libnvinfer_plugin.so.6: cannot open shared object file: No such file or directory
2020-07-13 19:17:41.866408: W tensorflow/compiler/tf2tensorrt/utils/py_utils.cc:30] Cannot dlopen some TensorRT libraries. If you would like to use Nvidia GPU with TensorRT, please make sure the missing libraries mentioned above are installed properly.

[skydoorkai]

This is a TensorFlow issue. TensorFlow will try to load TensorRT. If it is not installed, TensorFlow will generate these warnings.

TensorRT is optional. When present, TensorFlow may use it to accelerate GPU computation in graph mode.