7. Graph¶
7.1. What is a tf.Graph¶
In python, Graph is a class object, Class Graph.
TensorFlow provides a default graph that is an implicit argument(隐式参数) to all API functions in the same context.
tf.Graph包含两个相关联的信息：
7.1.1. Graph structure¶
7.1.2. Graph collections¶
- 本质就是一个dict，associate a list of objects with a key，就是说，一个key对应着一堆objects，这里的objects指的就是 tf.Operation objects and tf.Tensor objects
- 意义就是可以批量操作ops&tensors. Because disconnected parts of a TensorFlow program might want to create variables, it is sometimes useful to have a single way to access all of them。
在任意位置，任意层次都可以创造对象，存入相应collection中；创造完成后，统一从一个collection中取出一类变量，施加相应操作。
例如，tf.Variable()这个API就实现了上述两个信息：
- Executing v = tf.Variable(0) adds to the graph a tf.Operation that will store a writeable tensor value that persists between tf.Session.run calls.
- when you create a tf.Variable, it is added by default to collections representing “global variables” and “trainable variables”.
- tf.GraphKeys 包含了所有默认的集合名称。
- 实现方式就是新建ops object时，传入 collections 参数。
7.2. Executing a graph¶
我们构建的graph实际上就是一个client program，其和C++ runtime之间的连接由tf.Session来实现。
有三种方法执行一个图
- in a tf.Session.run(operation/tensor)
- operation.run()
- tensor.eval()