There are two simple ways to do this without using tensorboard, if you are using TPU VMs:
using tpu-info (read more here)
$ pip install git+https://github.com/google/cloud-accelerator-diagnostics/#subdirectory=tpu_info
$ tpu-info
Using Google Cloud Monitoring (metrics explorer) Go to this link and select which metrics you want to plot : https://console.cloud.google.com/monitoring/metrics-explorer