Tensorboard not found in google AI JupyterLab

Hi,
when i try to run tensorboard using the `google.datalab.ml` api, like this:

```python
from google.datalab.ml import TensorBoard
TensorBoard().start("gs://{}/flowers/trained_{}".format(BUCKET, MODEL_TYPE))
```

as used in many notebooks of the repo, in the new Google AI Notebook service, the tensorboard GUI is not reachable and it does not appear in the "tensorboards" tab.

I'm currently able to run it externally using the tensorboard bash command, but I'm wondering if there is a better way to solve the issue inside the notebooks.

Start by reproducing the `google.datalab.ml.TensorBoard` example in the affected Google AI Notebook service and check the notebook's "tensorboards" tab. Compare its behavior with the externally run TensorBoard bash command; done means the GUI is reachable from the notebook workflow or the supported alternative is clearly identified.

Written by the indexing model from the issue text.