我有一个训练有素的顺序模型，该模型由预先训练的无头高效网和最后一层组成。 model.summary()如下所示，
Model: "sequential"
_________________________________________________________________
Layer (type) Output Shape Param #
=================================================================
efficientnet-b3 (Model) (None, 5, 5, 1536) 10783528
_________________________________________________________________
gap (GlobalMaxPooling2D) (None, 1536) 0
_________________________________________________________________
dropout_out (Dropout) (None, 1536) 0
_________________________________________________________________
fc_out (Dense) (None, 1) 1537
=================================================================
Total params: 10,785,065
Trainable params: 1,479,937
Non-trainable params: 9,305,128
_________________________________________________________________
我尝试构建一个模型，该模型使用来输出efficiencynet-b3模型的预测输出和最后一个卷积层输出，
gradModel = Model(
inputs=[model.inputs],
outputs=[model.layers[0].layers[-3].output, model.output])
但是我得到了错误，
ValueError: Graph disconnected: cannot obtain value for tensor Tensor("input_1:0", shape=(None, 150, 150, 3), dtype=float32) at layer "input_1". The following previous layers were accessed without issue: []
我应该怎么解决这个问题？
我的efficiencynet-b3模型看起来像
Model: "efficientnet-b3"
__________________________________________________________________________________________________
Layer (type) Output Shape Param # Connected to
==================================================================================================
input_1 (InputLayer) [(None, 150, 150, 3) 0
__________________________________________________________________________________________________
conv2d (Conv2D) (None, 75, 75, 40) 1080 input_1[0][0]
__________________________________________________________________________________________________
.
.
.
__________________________________________________________________________________________________
add_18 (Add) (None, 5, 5, 384) 0 drop_connect_18[0][0]
batch_normalization_73[0][0]
__________________________________________________________________________________________________
conv2d_103 (Conv2D) (None, 5, 5, 1536) 589824 add_18[0][0]
__________________________________________________________________________________________________
batch_normalization_77 (BatchNo (None, 5, 5, 1536) 6144 conv2d_103[0][0]
__________________________________________________________________________________________________
swish_77 (Swish) (None, 5, 5, 1536) 0 batch_normalization_77[0][0]
==================================================================================================
Total params: 10,783,528
Trainable params: 1,478,400
Non-trainable params: 9,305,128
__________________________________________________________________________________________________