Tensorboard: Integrate hparams with global run selection

[wchargin]

Quoth @joshlk:

I've been using the HParams feature for a couple of days now and it
would still be really great to link the Tail ID to the Runs as
otherwise the HParams tab is totally disconnected from Scalars.

The HParams tab doesn't have the same overview of all runs like the
Scalars tab does. When I look at the results I start from the Scalars
tab and identify the runs that are doing well but then its very
difficult for me to find those runs in the HParams tab. Maybe it could
show the runs IDs when you select show metrics. Or you could filter by
run ID on the right side.

[vakker]

It would be a great feature to add.
Also, another solution could be to have all the values shown after clicking 'Show Metrics' on a single graph, not on individual graphs for each trial. It's very hard to compare metrics in it's current form.

[moritzmeister]

I finally had some time to get the HParams plugin integrated into our hyerparameter optimization framework. Great work @wchargin!

And I think the proposition of @vakker sounds like a good improvement. The scalar view and the HParams view are lacking a connection. I see a lot of people encoding the hyperparameters in folder names, but I think this shouldn't be the solution.

Bringing the hyperparameter information either into the tooltip for the global runs in the scalar tabs or introducing a global view in the Hparams plugin would be an awesome addition. Being able to select a subset of trials and showing them in a single graph would be great to compare metrics across trials.

In another iteration one could make the selection of runs in the scalar tab carry over to the hparams view. So say I identify a few runs in the scalar tab and select them, there should be a possibility to switch to hparams and show only the selected trails there, too.

How does the roadmap for improvements on the Hparams plugin look like? I might be able to contribute some spare time and work on it.