1cda66554188c4504aa94d1bcffd4cef9ee8a997,autokeras/net_transformer.py,,transform,#Any#,79

Before Change

    models = []
    layers = model.layers
    for index in range(len(layers) - 1):
        if isinstance(layers[index], tuple(WEIGHTED_LAYER_FUNC_LIST)):
            models.append(to_deeper_model(model, index))
            models.append(to_wider_model(model, index))
    return models

After Change

def transform(model):
    models = []
    conv_model = model.layers[0]
    dense_model = model.layers[1]

    for index, layer in enumerate(dense_model):
        // search dense variation
        if isinstance(layer, Dense):
            models.append(Sequential(copy_conv_model(conv_model), to_deeper_dense_model(dense_model, index)))
            models.append(Sequential(copy_conv_model(conv_model), to_wider_dense_model(dense_model, index)))

    for layer in conv_model.layers:
        // search conv variation
        if is_conv_layer(layer):
            models.append(Sequential(to_deeper_conv_model(conv_model, layer), dense_model))
            models.append(Sequential(to_wider_conv_model(conv_model, layer), dense_model))
    models.append(Sequential(to_skip_connection_model(conv_model), dense_model))

    return models

Instances