In TensorFlow, all computations are represented as graphs. A graph consists of nodes. The nodes in a graph are called operations (ops). An op or node can take tensors. Tensors are basically typed multi-dimensional arrays. For example, an image can be a tensor. So, in short, the TensorFlow graph has a description of all of the computation required.
In the preceding example, the ops of the graphs are as follows:
hello = tf.constant('Hello, TensorFlow!') a = tf.constant(12) b = tf.constant(34)
These tf.constant() methods create a constant op that will be added as a node in the graph. You can see how a string and integer are added to the graph.