HFAiLab/jupyterlab_tensorboard_pro

[Enhancement] Support s3 path as logdir

Closed this issue · 3 comments

Currently logdir is assumed to be a local FS path, but in a lot of cases, for example with distributed training, we may write tensorboard logs to s3.
Can we add support for passing s3 path as logdir and also make tensorboard capable of loading from s3 (by importing tensorflow_io)?

Thank you for your suggestion, I think it does make some sense.

I'll look into it this week and come up with an update plan.

Thanks! I think this feature could be really helpful

After a simple modification, the s3 path is now supported, and the README has been updated

In addition, I think some loading experience of tensorboard is not very good, I will continue to see how to optimize it later