Code to replicate performance metrics

[o1lo01ol1o]

Hi, This is a compelling library, do you have any of the code used to generate the reported performance increases over various frameworks mentioned here? I'm particularly curious about the tensorflow benchmark.

[wesm]

+1, the more easily reproducible the performance numbers the better (e.g. providing a Dockerfile would be ideal). This way users can validate performance on various hardware configurations

I also recommend for a project like this to do continuous performance testing. We use ASV, a Python library, in pandas to track operation microperformance so that we can catch performance regressions sooner.

[Lwelch45]

It would be interesting to see the steps taken for the spark integration.

[arjenpdevries]

Duplicate of #351