如何从已保存的模型加载权重

Question

I have a trained keras model saved with model.save(). When I load it and print a summary it appears as below.

Layer (type)                 Output Shape              Param #   
=================================================================
input_1 (InputLayer)         (None, 2)                 0         
_________________________________________________________________
model_1 (Model)              (None, 8)                 83208     
=================================================================
Total params: 83,208
Trainable params: 83,208
Non-trainable params: 0
_________________________________________________________________ 

Now I want to load the weights of the model_1 (a 6 layer network) in above network to a model with the same architecture. When I try to load from model.load_weights() I get an error saying that I cannot load from a 2 layer network to a 6 layer network. This is due to that model_1 is just a layer in the above model. How do I separately access and load weights from this layer?

Answer 1

You can use model.layers to access the various layers of a model and thus model.layers[1] to access model_1 . Then you can load the weights via model.layers[1].load_weights(...) .