melodyguan / enas

TensorFlow Code for paper "Efficient Neural Architecture Search via Parameter Sharing"
https://arxiv.org/abs/1802.03268
Apache License 2.0
1.58k stars 390 forks source link

Difference between fixed and searched path in reduction cells? #44

Open karandwivedi42 opened 6 years ago

karandwivedi42 commented 6 years ago

In this code:

            if self.fixed_arc is None:
              x = self._factorized_reduction(x, out_filters, 2, is_training)
              layers = [layers[-1], x]      # <<<<<<< HERE
              x = self._enas_layer(
                layer_id, layers, self.reduce_arc, out_filters)
            else:
              x = self._fixed_layer(
                layer_id, layers, self.reduce_arc, out_filters, 2, is_training,
                normal_or_reduction_cell="reduction")

On line 3, this gives layers=layers[-1], _factorised_reduction(layers[-1])]. Doesn't this make this inconsistent with the fixed cell?