pip install graphviz

import graphviz as gz

dot = gz.Digraph()
dot.node('1', 'Test1')
dot.node('2', 'Test2')
dot.node('3', 'Test3')
dot.node('4', 'Test4')
dot.edges(['12', '23', '34', '24'])
dot

dot = gz.Graph()
dot.node('1', 'Test1')
dot.node('2', 'Test2')
dot.node('3', 'Test3')
dot.node('4', 'Test4')
dot.edges(['12', '23', '34', '24'])
dot

import random

dot = gz.Digraph()
for i in range(10):
  dot.node('%s' % i, 'Test%s' % i)
dot.edges([str(random.randint(10, 99)) for i in range(10)])
dot

def neural_graph(inp=3, hide=(10, ), outp=3, inp_label='input', hide_label='hide', outp_label='output', dropout=True, style='h', size='2, 1'):
  """
  绘制简易神经网络图（有向图）
  :param inp: 输入神经元个数
  :param hide: 隐藏层神经元个数, 可迭代数组
  :param outp: 输出神经元个数
  :param inp_label: 输入名称显示
  :param hide_label: 隐藏层名称显示
  :param outp_label: 输出名称显示
  :param dropout: 是否全连接
  :param style: 水平或垂直显示， 可选项为 'h', 'v'
  :param size: 图像显示大小
  :return: 有向图
  """

  dot = gz.Digraph(name='neural network')
  dot.attr(size=size)
  if style == 'v':
    dot.attr(rankdir='LR')

  def draw(enter, exit, label1, label2):
    for i in range(enter):
      for j in range(exit):
        if dropout:
          if random.randint(0, max(enter, exit)):
            dot.edge('%s%s' % (label1, i), '%s%s' % (label2, j))
        else:
          dot.edge('%s%s' % (label1, i), '%s%s' % (label2, j))
  hide = list(hide)
  hide.insert(0, inp)
  hide.append(outp)
  for index, (i, j) in enumerate(zip(hide[:-1], hide[1:])):
    if index == 0:
      draw(i, j, inp_label, hide_label+str(index))
    elif index == len(hide) - 2:
      draw(i, j, hide_label+str(index-1), outp_label)
    else:
      draw(i, j, hide_label+str(index-1), hide_label+str(index))

  return dot

  #其他运行方式
  #return dot.view()