Tensorboard/en

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 the following icon on an active launcher tab: This will open the application on a new tab on 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 location where TensorBoard will look for data to display on your web browser, so you must make sure any calls to TensorBoard in your code write data to this directory. Failing to do so will result in no data being displayed on the TensorBoard tab on your browser. You can change the location of this directory by adding export TENSORBOARD_LOGDIR=/some/other/path in your .bashrc.

For detailed examples on the many uses of TensorBoard, see the official documentation for your preferred AI framework:

• PyTorch
• TensorFlow
• Jax