Tensorboardx: Projector dose not load new data

[mbenami]

Hi
I try to use the example for using the tensorboard projector as in the
demo_embedding.py but

but the tensorboard doesn't load new data
only if I stop it and relaunch it again (on the same folder ) I can see the new data

I'm trying to see the progress of the embedding cluster as the train goes

Thanks

[lanpa]

How do you add new data? Will the example work?
https://github.com/lanpa/tensorboardX/blob/master/examples/demo_multiple_embedding.py

[Timonzimm]

Same here, it doesn't refresh automatically and I have to restart tensorboard to get the new global_step that I added.

[LemonPi]

Can confirm this also happens when adding new runs.

[lanpa]

This requires a big rewrite to fix.

A quick fix would be adding return True in
~/anaconda3/envs/pytorch_dev/lib/python3.7/site-packages/tensorboard/plugins/projector/projector_plugin.py

def _latest_checkpoints_changed(configs, run_path_pairs):
  return True
  """Returns true if the latest checkpoint has changed in any of the runs."""

[mbenami]

Same here
tensorboard 2.2.1 tensorboardX (Version: 2.0+9d26706)
any update?
I try the workaround.. didn't help me

[MukundVarmaT]

Hey, is there any workaround for this? Restarting tensorboard does work but it's quite a hassle if I am running something on a remote server.