How to make the newly added Tensorboard in PyTorch 1.1 update with new data when it is opened inside a docker container?

How to make the newly added Tensorboard in PyTorch 1.1 update with new data when it is opened inside a docker container?

Currently, it does not update even if I press the refresh button on the top right corner. The only way to see the updated log is to restart the tensorboard instance. How to make it update without restarting the instance?

There is no problem if it is not run inside docker.

It seems like tensorflow is required in order for it to work in docker. I am not sure why. Anyone happens to know?