Figure out if we can reproduce benchmarks

[dfalbel]

A recent blog post in google AI has the following:

https://ai.googleblog.com/2021/12/interpretable-deep-learning-for-time.html