Support for gs:// and s3:// directories

[alexgorbachev]

Would it be possible to add a feature when gs:// or s3:// log_dir can be specified for Tensorboard to start? Assuming the environment is already authenticated to the GCP or AWS.
I'm running job in Google Cloud ML Engine and logs are in a GCS bucket. Also using JupyterLab inside a container on the cloud instances where this can be added too if supported by this package. chesterli29/jupyterlab_tensorboard#9 (comment)
cc @chaoleili

[diogoseca]

While this is still not working, you can start tensorboard from the Cloud Shell, and preview on port 6006 (or whatever port you use for running tensorboard).

However, it'd be nice to open tensorboard from the jupyterlab without this workaround.

[wookayin]

+1