Permalink
Switch branches/tags
Nothing to show
Find file Copy path
769033a Sep 28, 2018
2 contributors

Users who have contributed to this file

@qfyin @kant
45 lines (24 sloc) 1.64 KB

Run TensorBoard locally in VS Code

In this quickstart, we will run a tensorboard service locally in AI Tools and view the model graph.

Prerequisites

Before you begin, ensure you set Python path properly in VS Code properly and have already installed following packages in Python:

Google TensorFlow

Install tensorflow in the python path with pip

pip install tensorflow

or if you have an Nvidia GPU

pip install tensorflow-gpu

For more info, please refer to official TensorFlow doc. The TensorBoard is contained in TensorFlow package, so you don't need to install TensorBoard separately.

Write the log data when you training a TensorFlow job.

Use tf.summary module to collect data from your training job. Here is an example project from TensorBoard official guide.

Select the log directory and start TensorBoard from VS Code command palette.

  • Open command palette from VS Code by pressing Ctrl+Shift+P.

  • Select "AI: Local - Run TensorBoard" command

Command Palette

  • A message window will pop up for selecting the log directory created by training job.

select log directory

  • After selecting the log directory, a TensorBoard server will start in the VS Code terminal and represent the model graph in an external browser.

show tensorboard in browser