GPU Support

Known issues with GPU support.

Only CUDA-enabled NVIDIA GPU hardware is supported

Cloudera Data Science Workbench only supports CUDA-enabled NVIDIA GPU cards.

Heterogeneous GPU hardware is not supported

You must use the same GPU hardware across a single Cloudera Data Science Workbench deployment.

GPU image for CDSW does not work with TensorFlow

The LD_LIBRARY_PATH environment variable is not set properly in the technical preview GPU image (docker.repository.cloudera.com/cdsw/cuda-engine:10) which is needed for the TensorFlow framework to work.

Workaround: To use the technical preview GPU image (docker.repository.cloudera.com/cdsw/cuda-engine:10) with TensorFlow:
  1. Install TensorFlow by running the following command:
    pip3 install tensorflow
  2. Add the following to the LD_LIBRARY_PATH environment variable:
    LD_LIBRARY_PATH=/usr/local/cuda/lib64:/usr/local/cuda/lib:/usr/lib/hadoop/lib/native