stackeddag

A visualization tool to show a ascii graph from Graphviz-Dot-file or Tensorflow


Keywords
tensorflow, tensor, machine-learning, graphviz, ascii, dag, ml, deep-learning, neural-network
License
MIT
Install
pip install stackeddag==0.3.4

Documentation

stacked-dag for python

Build Status

Ascii DAG for visualization of dataflow

stacked-dag can show Ascii-DAG(Directed acyclic graph) from a Dot file of graphviz. Dataflow's direction is from top to bottom. 'o' means a node. A label of the node is shown to the left side. Other characters are edges of DAG.

A sample of DAG is below.

o o    l0,l4
|/
o    l1
|
o    l2
|
o    l3

Usage with dot

Write a Dot file of graphviz.

$ cat > sample.dot
digraph graphname {
  0 [label="l0"];
  1 [label="l1"];
  2 [label="l2"];
  3 [label="l3"];
  4 [label="l4"];
  0->1;
  1->2;
  2->3;
  4->1;
}

Show ascii DAG by following command.

$ python stackeddag sample.dot
o o    l0,l4
|/
o    l1
|
o    l2
|
o    l3

Usage with tensorflow

import tensorflow as tf
import stackeddag.tf as sd

def mydataflow():
  a = tf.constant(1,name="a")
  b = tf.constant(2,name="b")
  c = tf.add(a,b,name="c")
  return tf.get_default_graph()

print(sd.fromGraph(mydataflow()),end="")

The output is below.

o o    a,b
|/
o    c

Another sample

$ python stackeddag sample/test.dot
o
|
o
|\
o |
| |\
o o |
|\ \ \
| | |\ \
| | | | |\
o o o o o |
|/ /_/ / /
| |  / /
o o o o
|/_/_/
o