"add net drawer for visualizing the graph"

[dzhwinter]

add a visualizer for drawing the ProgramDesc.
It's very simple for debugging your python program.
usage:

import paddle.v2.framework.net_drawer as drawer
print drawer.draw_graph(init_program, program)

then you will get the

digraph <function generator at 0x7f50ba8bfe60> {
	graph [rankdir=TB]
	node [color="#0F9D58" fontcolor="#FFFFFF" shape=oval style=filled]
	fill_constant [label=fill_constant color="#0F9D58" fontcolor="#FFFFFF" shape=oval style=filled]
	uniform_random [label=uniform_random color="#0F9D58" fontcolor="#FFFFFF" shape=oval style=filled]
	mul [label=mul color="#0F9D58" fontcolor="#FFFFFF" shape=oval style=filled]
	Feed -> mul [label="X(x)"]
	uniform_random -> mul [label="Y(fc_0.w_0)"]
	elementwise_add [label=elementwise_add color="#0F9D58" fontcolor="#FFFFFF" shape=oval style=filled]
	mul -> elementwise_add [label="X(fc_0.tmp_0)"]
	fill_constant -> elementwise_add [label="Y(fc_0.b_0)"]
	elementwise_sub [label=elementwise_sub color="#0F9D58" fontcolor="#FFFFFF" shape=oval style=filled]
	elementwise_add -> elementwise_sub [label="X(fc_0.tmp_1)"]
	Feed -> elementwise_sub [label="Y(y)"]
	square [label=square color="#0F9D58" fontcolor="#FFFFFF" shape=oval style=filled]
	elementwise_sub -> square [label="X(square_error_cost_0.tmp_0)"]
	mean [label=mean color="#0F9D58" fontcolor="#FFFFFF" shape=oval style=filled]
	square -> mean [label="X(square_error_cost_0.tmp_1)"]
}

which can be shown in a web browser.

[tonyyang-svail]

Looks good to me.

[dzhwinter]

We will enhance the RNN and other blocks supporting soon.