Question

我有以下代码将DNN模型分为两部分。

def split(model, input):
starting_layer_name = input
new_input = layers.Input( batch_shape=model.get_layer(starting_layer_name).get_input_shape_at(0))
layer_outputs = {}

def get_output_of_layer(layer):

    if layer.name in layer_outputs:
        return layer_outputs[layer.name]

    if layer.name == starting_layer_name:
        out = layer( new_input )
        layer_outputs[layer.name] = out
        return out
    prev_layers = []

    for node in layer._inbound_nodes:
        prev_layers.extend( node.inbound_layers )

    pl_outs = []

    for pl in prev_layers:
        pl_outs.extend( [get_output_of_layer( pl )] )

    out = layer( pl_outs[0] if len( pl_outs ) == 1 else pl_outs )
    layer_outputs[layer.name] = out
    return out

if starting_layer_name=='input_1':
    new_output = get_output_of_layer(model.layers[-21])
    block_1 = models.Model( new_input, new_output )
    return block_1

elif starting_layer_name=='block1_pool':
    new_output =get_output_of_layer((model.layers[-1]))
    block_2 = models.Model(new_input, new_output)
    return block_2

block_1=split(model,'input_1')
block_2=split(model,'block1_pool')

block_1.save('my_model1.h5')
block_2.save('my_model2.h5')

当我尝试运行以下代码时，我检索到“图形断开连接无法获取张量值。

from Keras.models import load_model

model = load_model('my_model1.h5')
model.summary()

在解决此问题方面提供帮助非常感谢。我正在尝试拆分模型的当前方法给我一个错误，是否有另一种方法可以解决keras中的此问题。

Answer 1

我已经更新了分割点的代码，现在load方法可以正常工作了。请在下面找到更新的代码

def split_keras_model(model, index):

 layer_input_1 = Input(model.layers[0].input_shape[1:])
 x = layer_input_1
 # Foreach layer: connect it to the new model
 for layer in model.layers[1:index]:
    x = layer(x)
 model1 = Model(inputs=layer_input_1, outputs=x)
 input_shape_2 = model.layers[index].get_input_shape_at(0)[1:]
 print("Input shape of model 2: " + str(input_shape_2))
 # A new input tensor to be able to feed the desired layer
 layer_input_2 = Input(shape=input_shape_2)
 x = layer_input_2
 for layer in model.layers[index:]:
    x = layer(x)
 model2 = Model(inputs=layer_input_2, outputs=x)

 return (model1, model2)

model1,model2=split_keras_model(model,2)

model1.save('my_model1.h5')
loaded_model=load_model('my_model1.h5',compile=False)

加载预训练模型VGG-16的问题

1 个答案: