More examples - Gluon for hybrid block


Hi There

I am trying to implement text classification using CNNs. Gluon tutorials only consist of models being initialized as Net = gluon.nn.Sequential(). I want to see examples which makes use of block or hybrid block class. Could you please redirect me where I can find such examples ?



A question I have for HybridBlocks is how to create a “variable” number of hidden layers.

If I try something like:

self.layer_sizes = kwargs.pop('layer_sizes', [512,512])

with self.name_scope():
   for layer in self.layer_sizes:
       self.fcs = [nn.Dense(layer) for layer in self.layer_sizes]

It tells me that the weight parameter has not been initialized.