1) Removed PyTorch samples. 2) Updated README files. 3) Verified that all the training scripts are upto date with those available in the corresponding framework repos.

1) Combined updates from jureca_2019_a and juwels_2019a. 2) Removed python 2 job submission scripts. 3) Started making other required updates.

Upon deciding to remove sample scripts for Python 2, removed the python versions from the script names. Also, all sample job submission scripts for pytorch have been hidden, and PyTorch examples do not work on JUWELS; with or without Horovod.

For PyTorch samples (with and without Horovod), the manner in which pre-downloaded datasets are loaded has been changed a bit to comply with the versions of torch and torchvision installed in stage 2019a.

Added job scripts for JUWELS stage 2019a. Keras and Tensorflow are working both independently and with Horovod. PyTorch code is generating dataset related errors that need to be investigated; possibly just a matter of updating the custom pre-downloaded-dataset handling code.

JURECA job scripts now load the 2018b rather than using the default stage, as the DL frameworks are only available on 2018b.

The git-lfs module for JURECA is not available in 2019a. Therefore, a new step has been added where the previous stage is loaded for cloning this repository.

Add table of contents

Closes #1

See merge request !1

JURECA job scripts for Keras, Tensorflow, PyTorch, and Horvood now load modules from the production stage.

Bug fix: The 'run_on_localMachine.sh' script in the 'keras' directory now uses the correct file name.