#!/usr/bin/env python
# coding: utf-8

# In[3]:


# %load /Users/facai/Study/book_notes/preconfig.py
get_ipython().run_line_magic('matplotlib', 'inline')

import matplotlib.pyplot as plt
import seaborn as sns
sns.set(color_codes=True)
sns.set(font='SimHei', font_scale=2.5)
plt.rcParams['axes.grid'] = False

import tensorflow as tf

def show_image(filename, figsize=None, res_dir=True):
    if figsize:
        plt.figure(figsize=figsize)

    if res_dir:
        filename = './res/{}'.format(filename)

    plt.imshow(plt.imread(filename))


# TensorFlow图相关知识简介
# ============
# 
# 参考: https://www.tensorflow.org/programmers_guide/graphs
# 
# 1. tf.Graph
# 
# 2. op, tensor
# 
# 3. variable
# 
# 4. name_scope, variable_scop, collection
# 
# 5. save and restore

# ![](https://www.tensorflow.org/images/tensors_flowing.gif)

# ### 0. tf.Graph
# 
# tf.Graph: GraphDef => *.pb文件
# + Graph structure: Operator, Tensor-like object, 连接关系
# + Graph collections: metadata
# 
# tf.Session():
# + 本地
# + 分布式：master (worker_0)
# 
# ```python
# with tf.Session("grpc://example.org:2222"):
#   pass
# ```
# 
# 状态（Variable) => *.ckpt文件

# ### 1. 算子与Tensor

# In[6]:


a = tf.constant(1)
b = a * 2
b


# In[7]:


b.op


# In[11]:


b.consumers()


# In[15]:


a.op


# In[19]:


a.consumers()


# tensorflow/python/framework/ops.py
# + Tensor:
#     - device
#     - graph
#     - op
#     - consumers
#     - _override_operator:
#       数学算子：math_op.add 重载 `__add__`

# In[8]:


b.op.outputs


# In[9]:


list(b.op.inputs)


# In[14]:


print(b.op.inputs[0])
print(a)


# In[17]:


list(a.op.inputs)


# + Operator: NodeDef
#    - device
#    - inputs
#    - outputs
#    - graph
#    - node_def
#    - op_def
#    - **run**
#    - traceback
#    
# Operator和Tensor构成无向图
# 
# ```python
# # run
# sess.run([b])
# ```
# 
# 参考:
# + tf.Tensor: https://www.tensorflow.org/versions/master/api_docs/python/tf/Tensor
# + tf.Operator: https://www.tensorflow.org/versions/master/api_docs/python/tf/Operation

# ### 2. 变量

# In[20]:


v = tf.Variable([0])
c = b + v
c


# In[23]:


list(c.op.inputs)


# In[25]:


c.op.inputs[1].op


# In[26]:


list(c.op.inputs[1].op.inputs)


# In[21]:


v


# 实际上，对变量的读是通过`tf.identity`算子得到：
# ```python
# c = tf.add(b, tf.identity(v))
# ```
# 
# 
# + Variable: act like Tensor
#    - [ops](https://stackoverflow.com/questions/40817665/whats-the-difference-between-variable-and-resourcevariable-in-tensorflow)
#      1. VariableV2
#      2. ResourceVariable
#    - _AsTensor -> g.as_graph_element
#    - value: Identity(variable) -> Tensor
#    - assign
#    - init_op: Assign(self, init_value)
#    - to_proto: VariableDef
# 
# 参考：https://www.tensorflow.org/versions/master/api_docs/python/tf/Variable

# 3. collections
# ===============
# 
# + collections: 按作用分组
#   - Variable: global_varialbe
#   - 更多见[tf.GraphKeys](https://www.tensorflow.org/versions/master/api_docs/python/tf/GraphKeys)
# + name_scope: Operator, Tensor
# + variable_scope: Variable
#   - 伴生name_scope
#   
# ```python
# class Layer:
#   def build(self):
#     pass
#   def call(self, inputs):
#     pass
# ```
# 
# ![](https://www.tensorflow.org/images/mnist_deep.png)
# 
# 参考：https://www.tensorflow.org/programmers_guide/summaries_and_tensorboard

# ### 4. 保存与恢复

# In[58]:


graph_a = tf.Graph()
with graph_a.as_default():
    v1 = tf.get_variable("v1", shape=[3], initializer = tf.zeros_initializer)
    print(v1)
    inc_v1 = v1.assign(v1+1)
      
    init_op = tf.global_variables_initializer()
    saver = tf.train.Saver()
  
    with tf.Session() as sess:
        sess.run(init_op)
        inc_v1.op.run()
        save_path = saver.save(sess, "./tmp/model.ckpt", write_meta_graph=True)
        print("Model saved in path: %s" % save_path)
        
        pb_path = tf.train.write_graph(graph_a.as_graph_def(), "./tmp/", "graph.pbtxt", as_text=True)
        print("Graph saved in path: %s" % pb_path)


# graph.pbtxt部份示意：`v1 + 1`:
# 
# ```bash
# node {
#   name: "add"
#   op: "Add"
#   input: "v1/read"
#   input: "add/y"
#   attr {
#     key: "T"
#     value {
#       type: DT_FLOAT
#     }
#   }
# }
# ```

# In[62]:


graph_b = tf.Graph()
with graph_b.as_default():
    with tf.Session() as sess:
        saver = tf.train.import_meta_graph('./tmp/model.ckpt.meta')
        saver.restore(sess, "./tmp/model.ckpt")
        print(graph_b.get_operations())
        
        v1 = graph_b.get_tensor_by_name("v1:0")
        print("------------------")
        print("v1 : %s" % v1.eval(session=sess))


# 总结：
# + `tf.train.Saver`会保存GraphDef和Variable信息，用它可以直接恢复图。
#   - [tf.train.import_meta_graph](https://www.tensorflow.org/versions/master/api_docs/python/tf/train/import_meta_graph)
#   - [Exporting and Importing a MetaGraph](https://www.tensorflow.org/versions/master/api_guides/python/meta_graph)
#   - 缺点：无法指input_tensor。
# + `tf.train.write_graph`、`tf.GraphDef`和`tf.import_graph_def`，主要用于固化模型（只有GraphDef信息）。
# 
# 
# 参考：
# + https://stackoverflow.com/questions/38641887/how-to-save-a-trained-tensorflow-model-for-later-use-for-application
# + https://www.tensorflow.org/programmers_guide/saved_model#overview_of_saving_and_restoring_models

# In[ ]: