b048efa4f956b80266942caea6ee8b4311e1d17a,autokeras/net_transformer.py,,to_skip_connection_graph,#Any#,57

Before Change

        The skip_connected model
    
    // The last conv layer cannot be widen since wider operator cannot be done over the two sides of flatten.
    weighted_layers = list(filter(lambda x: is_conv_layer(x), graph.layer_list))[:-1]
    index_a = randint(0, len(weighted_layers) - 1)
    index_b = randint(0, len(weighted_layers) - 1)
    if index_a == index_b:
        if index_b == 0:
            index_a = index_b + 1
        else:
            index_a = index_b - 1
    if index_a > index_b:
        index_a, index_b = index_b, index_a
    a = weighted_layers[index_a]
    b = weighted_layers[index_b]
    a_id = graph.layer_to_id[a]
    b_id = graph.layer_to_id[b]
    if a.output_shape[-1] != b.output_shape[-1]:
        graph.to_concat_skip_model(a_id, b_id)
    elif random() < 0.5:
        graph.to_add_skip_model(a_id, b_id)

After Change

        The skip_connected model
    
    // The last conv layer cannot be widen since wider operator cannot be done over the two sides of flatten.
    weighted_layer_ids = graph.skip_connection_layer_ids()
    index_a = randint(0, len(weighted_layer_ids) - 1)
    index_b = randint(0, len(weighted_layer_ids) - 1)
    if index_a == index_b:
        if index_b == 0:
            index_a = index_b + 1
        else:
            index_a = index_b - 1
    if index_a > index_b:
        index_a, index_b = index_b, index_a
    a_id = weighted_layer_ids[index_a]
    b_id = weighted_layer_ids[index_b]
    if graph.layer_list[a_id].output_shape[-1] != graph.layer_list[b_id].output_shape[-1]:
        graph.to_concat_skip_model(a_id, b_id)
    elif random() < 0.5:
        graph.to_add_skip_model(a_id, b_id)

Instances