TensorFlow Debugger

TDBCame across this very handy tool for visualizing progress when training a TensorFlow graph. The screen capture shows the demo Jupyter Notebook for the mnist_demo. It’s pretty easy to set up, assuming TensorFlow has been installed…

It’s necessary to install Jupyter if not already installed:

sudo pip install jupyter

Also, install tfdebugger:

sudo pip install tfdebugger

Then, start Python in a terminal window and:

import notebook.nbextensions
import urllib
import zipfile
SOURCE_URL = 'https://github.com/ericjang/tdb/releases/download/tdb_ext_v0.1/tdb_ext.zip'
urllib.urlretrieve(SOURCE_URL, 'tdb_ext.zip')
with zipfile.ZipFile('tdb_ext.zip', "r") as z:
    z.extractall("")
notebook.nbextensions.install_nbextension('tdb_ext',user=True)

The Python session can then be terminated. Using the terminal again:

cd <path to tensorflow>
git clone https://github.com/ericjang/tdb.git
jupyter notebook

This will start the server running and also open up a browser window. Use the url localhost:8888 to connect to the Jupyter server. On the files tab navigate to tdb/notebooks and click on mnist_demo.ipynb.  Then run the notebook and watch the learning progress!

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s