@rgommers: any opinion on exactly what version we should use as the upper bound? I just chose 60.0 here, but saw just now that numpy uses a specific version (==59.2.0)
I would suggest <65.0. The 64.x versions seem to work, and perhaps have some relevant fixes. Many versions will likely do, this just seems like a logical choice.
closes #645
@rgommers: any opinion on exactly what version we should use as the upper bound? I just chose 60.0 here, but saw just now that numpy uses a specific version (
==59.2.0
)