Add option to use display/friendly names for runs in Tensorboard

[tomasvanpottelbergh]

When using the Tensorboard class of the Azure ML SDK, runs are named after their Run ID. This makes it hard to compare runs, as you need to remember what the differences are between the IDs.

It would be very much appreciated if you could add an option to use the Display name (or allow to set a custom name) for the runs in Tensorboard instead.

[v-kumudam]

@tomasvanpottelbergh Is the issue (Display name (or allow to set a custom name) for the runs in Tensor board) is resolved?

[tomasvanpottelbergh]

@v-kumudam No, I didn't see an update of the azureml-tensorboard package.

[v-kumudam]

"run = exp.submit(src)\n",
"# You can also wait for the run to complete\n",
"# run.wait_for_completion(show_output=True)\n",
"runs.display_name = 'Tensorboard_Demo'\n",
"runs.append(run)"

We can solve this problem by adding this line..

[tomasvanpottelbergh]

I don't understand what you mean. My feature request is to add an option to the Tensorboard class constructer, for example:
Tensorboard(runs, local_root=None, port=6006, use_display_names=False).

Setting use_display_names=True would then mount the TensorBoard logs to folders named after the run.display_name instead of run.id (the current behaviour).

The issue for me is the naming of the runs as displayed in TensorBoard, when using the Azure ML Tensorboard class. I hope this clarifies what I mean?