Running TensorBoard is really great tool for visualizing your TensorFlow model and training workflow. To get it started in Paperspace is really simple!
1. Launch a TensorFlow notebook
Let's launch a new notebook using the Tensorflow 1.5.0 Py3 base container. If you are using a container that does not already include Tensorboard you can easily install it from within the notebook using pip install tensorboard but in this case the container already includes tensorboard.
1. Open a terminal from within the notebook
Once the notebook is running, click Open to go to the public IP address and see your running notebook instance. From here you need to open a new Terminal by clicking New → Terminal
3. Start Tensorboard
Type tensorboard --logdir logs where logsdir is where you will be pointing your TensorFlow codes to.
4. Open the public IP address on port 6006
If our notebook is running on 35.320.103.68 then open a new web browser tab and go to that IP address at port 6006. You should now see TensorBoard running!