TensorFlow2.x中的session.run()等效项

Question

此函数采用预先训练好的imagenet-vgg-verydeep-19.mat模型，并使用其权重和输入返回图( model )。

def load_vgg_model(path):
    ...
    ...
    ...
    # Constructs the graph model.
    graph = {}
    graph['input']   = tf.Variable(np.zeros((1, 400, 300, 3)), dtype = 'float32')
    graph['conv1_1']  = _conv2d_relu(graph['input'], 0, 'conv1_1')      # returns tf.nn.relu
    graph['conv1_2']  = _conv2d_relu(graph['conv1_1'], 2, 'conv1_2')
    graph['avgpool1'] = _avgpool(graph['conv1_2'])                    # returns tf.nn.avg_pool
    graph['conv2_1']  = _conv2d_relu(graph['avgpool1'], 5, 'conv2_1')
    ....
    ....
    graph['conv5_4']  = _conv2d_relu(graph['conv5_3'], 34, 'conv5_4')
    graph['avgpool5'] = _avgpool(graph['conv5_4'])
    
    return graph

下面是TensorFlow 1中相同的实现

model = load_vgg_model("imagenet-vgg-verydeep-19.mat")
sess = tf.compat.v1.InteractiveSession()
content_image = imread("images/louvre_small.jpg")

sess.run(model['input'].assign(content_image))
out = model['conv4_2']
a_C = sess.run(out)
a_G = out

但是我想知道TensorFlow 2.x中的实现

我已经阅读了这些文档https://www.tensorflow.org/guide/effective_tf2

https://www.tensorflow.org/guide/migrate

Answer 1

正如在this doc中提到的，Tensorflow 2具有默认的紧急执行。代码通过构建中间图来执行。

在Tensorflow 2中，会话的创建不是required.You，可以使用Autograph函数来代替会话来创建图形。

文件上说，

TensorFlow 1.x requires users to manually stitch together an abstract syntax tree (the graph) by making tf.* API calls. It then requires users to manually compile the abstract syntax tree by passing a set of output tensors and input tensors to a session.run call. TensorFlow 2.0 executes eagerly (like Python normally does) and in 2.0, graphs and sessions should feel like implementation details.

One notable byproduct of eager execution is that tf.control_dependencies is no longer required, as all lines of code execute in order (within a tf.function, code with side effects executes in the order written).