TensorBoard¶

TensorBoard provides tools for visualizing and monitoring machine learning experiments. It is often used by TensorFlow, Keras and PyTorch users.

Availability on Cori¶

You can run TensorBoard client in your browser from https://jupyter.nersc.gov/. In a separate notebook, choose any TensorFlow jupyter kernel then run:

import nersc_tensorboard_helper
%load_ext tensorboard

Run a tensorboard server from Jupyter by running the following command in a new cell (note that port 0 asks TensorBoard to use a port not already in use):

%tensorboard --logdir YOURLOGDIR --port 0

Calling the NERSC TB helper function will provide you with an address to connect to the TensorBoard instance:

nersc_tensorboard_helper.tb_address()