Add Nvidia DALI tutorial

[axkoenig]

Description

Summary

• Integrated Squirrel and DALI by defining an "external source" for a DALI Pipeline (read more here)
• Why do that? DALI is a library that is highly optimised for computations on the GPU. With this integration we get very fast image augmentations and other image processing that DALI offers, right out of the box.
• Tested on cloud machine with 1 Nvidia Tesla T4
• This tutorial is based on the squirrel Catalog API. For the benchmarks, we used a local MessagePack copy of the CIFAR dataset instead, which was much faster than the Catalog API.

Required dependencies (also mentioned in code)

pip install cupy-cuda113 squirrel-core squirrel-datasets-core
pip install --extra-index-url https://developer.download.nvidia.com/compute/redist --upgrade nvidia-dali-cuda110

Script prints

Building pipeline ...
Iterating over dataset ...
Data in torch format!
Iteration speed: 2213.6659146588822 samples per second
Done. Have a nice day.

Type of change

• Bug fix (non-breaking change which fixes an issue)
• New feature (non-breaking change which adds functionality)
• Breaking change (fix or feature that would cause existing functionality to not work as expected)
• Documentation update
• Refactoring including code style reformatting
• Other (please describe):

Checklist:

• I have read the contributing guideline doc (external contributors only)
• Lint and unit tests pass locally with my changes
• I have kept the PR small so that it can be easily reviewed
• I have made corresponding changes to the documentation
• I have added tests that prove my fix is effective or that my feature works
• All dependency changes have been reflected in the pip requirement files.

[winfried-ripken]

Thanks Alex!