Tensorboard
TensorBoard is a suite of web applications for inspecting and understanding your AI and Machine Learning runs. It includes visual tools to track performance and evaluation metrics, profile code, explore intermediate layers inside models, visualize embeddings and more. Originally built for TensorFlow, it also supports other frameworks such as PyTorch and Jax.
On JupyterHub¶
On clusters where JupyterHub is available, you can launch TensorBoard by clicking on an icon on an active launcher tab. This will open the application in a new tab in your web browser. Switch to that tab to start using TensorBoard.
TensorBoard Log Directory
Upon launching TensorBoard on JupyterHub, a directory $HOME/tensorboard_logs will be created. This is the default location where TensorBoard will look for data to display in your web browser, so you must ensure any calls to TensorBoard in your code write data to this directory. Failing to do so will result in no data being displayed in the TensorBoard tab in your browser.
You can change the default location of this directory by adding the following to your .bashrc file:
For detailed examples on the many uses of TensorBoard, see the official documentation for your preferred AI framework: