[Question] Meta Learning Lecture: get_convnet() output_size

[jeffwiroj]

I am perhaps misunderstanding the get_convnet function, but if the output_size corresponds to the num_classes, why would we set it to the proto_dim? I don't get this line: self.model = get_convnet(output_size=self.hparams.proto_dim). Shouldn't we set the output_size to N_way, which is the number of classes. Right now we are setting it to the proto_dim=64, which far exceeds the number of classes.