A simple graphviz wrapper to visualize a graph like TensorFlow and TensorBoard
tfgraphviz is a module to create a TensorBoard-ish graph in the DOT language of the Graphviz. tfgraphviz provides a visualization of tensorflow graph on Jupyter Notebook without TensorBoard.
- GitHub: http://github.com/akimach/tfgraphviz
- PyPI: https://pypi.python.org/pypi/tfgraphviz
- Jupyter Notebook: https://nbviewer.jupyter.org/github/akimach/tfgraphviz/blob/master/examples/jupyter_sample.ipynb
Use pip to install:
$ pip install tfgraphvizThe only dependency is Graphviz.
import tensorflow as tf
import tfgraphviz as tfg
a = tf.constant(1, name="a")
b = tf.constant(2, name="b")
c = tf.add(a, b, name="add")
g = tfg.board(tf.get_default_graph())
g.view()
This package is distributed under the MIT license.
Akimasa KIMURA