[Python] how to easily modify the number of neurons of a certain layer in a neural network...

In tensorflow, given a model h = Model(input_layer, output_layer) for example (its summary looks like below)

Model: "functional_1"
┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type) ┃ Output Shape ┃ Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer (InputLayer) │ (None, 784) │ 0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense (Dense) │ (None, 64) │ 50,240 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_1 (Dense) │ (None, 32) │ 2,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_2 (Dense) │ (None, 64) │ 2,112 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_3 (Dense) │ (None, 784) │ 50,960 │
└─────────────────────────────────┴────────────────────────┴───────────────┘
Total params: 105,392 (411.69 KB)
Trainable params: 105,392 (411.69 KB)
Non-trainable params: 0 (0.00 B)


If I want to change the number of neurons of dense_1 from 32 to 16, is there any easy and handy way to update the new configuration?

I tried h.layers[2].units = 16, which unfortunately didn't change the model setup at all. I am wondering if there is kind of "assignment" operation to do that.

Continue reading...