Time major data layout for RNN

This example demonstrates an RNN implementation with Time-major layout. This implementation shows 1.5x-2x speedups compared to Batch-major RNN.

As example of Batch-major RNN is available in MXNet RNN Bucketing example

Running the example

• Prerequisite: an instance with GPU compute resources is required to run MXNet RNN

• Make the shell script get_ptb_data.sh executable:

  chmod +x get_ptb_data.sh
  

• Run get_ptb_data.sh to download the PTB dataset, and follow the instructions to review the license:

  ./get_ptb_data.sh
  

  The PTB data sets will be downloaded into ./data directory, and available for the example to train on.

• Run the example:

  python python rnn_cell_demo.py
  

  If everything goes well, console will plot training speed and perplexity that you can compare to the batch major RNN.