rename and export call_layer to create_layer

anuj2054 · Jul 14, 2017 · e697562 · e697562
1 parent 0cbad8b
commit e697562
Show file tree

Hide file tree

Showing 13 changed files with 102 additions and 61 deletions.
diff --git a/NAMESPACE b/NAMESPACE
@@ -40,6 +40,7 @@ export(constraint_minmaxnorm)
 export(constraint_nonneg)
 export(constraint_unitnorm)
 export(count_params)
+export(create_layer)
 export(dataset_boston_housing)
 export(dataset_cifar10)
 export(dataset_cifar100)

diff --git a/R/layer-wrappers.R b/R/layer-wrappers.R
@@ -21,7 +21,7 @@ time_distributed <- function(object, layer, input_shape = NULL,
                              batch_input_shape = NULL, batch_size = NULL, dtype = NULL, 
                              name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$TimeDistributed, object, list(
+  create_layer(keras$layers$TimeDistributed, object, list(
     layer = layer,
     input_shape = normalize_shape(input_shape),
     batch_input_shape = normalize_shape(batch_input_shape),
@@ -51,7 +51,7 @@ bidirectional <- function(object, layer, merge_mode = "concat", input_shape = NU
                           batch_input_shape = NULL, batch_size = NULL, dtype = NULL, 
                           name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$Bidirectional, object, list(
+  create_layer(keras$layers$Bidirectional, object, list(
     layer = layer,
     merge_mode = merge_mode,
     input_shape = normalize_shape(input_shape),

diff --git a/R/layers-activations.R b/R/layers-activations.R
@@ -15,7 +15,7 @@ layer_activation <- function(object, activation, input_shape = NULL,
                              batch_input_shape = NULL, batch_size = NULL, dtype = NULL, 
                              name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$Activation, object, list(
+  create_layer(keras$layers$Activation, object, list(
     activation = activation,
     input_shape = normalize_shape(input_shape),
     batch_input_shape = normalize_shape(batch_input_shape),
@@ -47,7 +47,7 @@ layer_activation_leaky_relu <- function(object, alpha = 0.3, input_shape = NULL,
                                         dtype = NULL, name = NULL, trainable = NULL, 
                                         weights = NULL) {
 
-  call_layer(keras$layers$LeakyReLU, object, list(
+  create_layer(keras$layers$LeakyReLU, object, list(
     alpha = alpha,
     input_shape = normalize_shape(input_shape),
     batch_input_shape = normalize_shape(batch_input_shape),
@@ -105,7 +105,7 @@ layer_activation_parametric_relu <- function(object, alpha_initializer = "zeros"
   args$weights <- weights
 
   # call layer
-  call_layer(keras$layers$PReLU, object, args)
+  create_layer(keras$layers$PReLU, object, args)
 }
 
 
@@ -126,7 +126,7 @@ layer_activation_thresholded_relu <- function(object, theta = 1.0, input_shape =
                                               dtype = NULL, name = NULL, trainable = NULL, 
                                               weights = NULL) {
 
-  call_layer(keras$layers$ThresholdedReLU, object, list(
+  create_layer(keras$layers$ThresholdedReLU, object, list(
     theta = theta,
     input_shape = normalize_shape(input_shape),
     batch_input_shape = normalize_shape(batch_input_shape),
@@ -158,7 +158,7 @@ layer_activation_elu <- function(object, alpha = 1.0, input_shape = NULL,
                                  batch_input_shape = NULL, batch_size = NULL, dtype = NULL, 
                                  name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$ELU, object, list(
+  create_layer(keras$layers$ELU, object, list(
     alpha = alpha,
     input_shape = normalize_shape(input_shape),
     batch_input_shape = normalize_shape(batch_input_shape),

diff --git a/R/layers-convolutional.R b/R/layers-convolutional.R
@@ -56,7 +56,7 @@ layer_conv_1d <- function(object, filters, kernel_size, strides = 1L, padding =
                           batch_input_shape = NULL, batch_size = NULL, dtype = NULL, 
                           name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$Conv1D, object, list(
+  create_layer(keras$layers$Conv1D, object, list(
     filters = as.integer(filters),
     kernel_size = as_integer_tuple(kernel_size),
     strides = as_integer_tuple(strides),
@@ -138,7 +138,7 @@ layer_conv_2d <- function(object, filters, kernel_size, strides = c(1L, 1L), pad
                           batch_input_shape = NULL, batch_size = NULL, dtype = NULL, 
                           name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$Conv2D, object, list(
+  create_layer(keras$layers$Conv2D, object, list(
     filters = as.integer(filters),
     kernel_size = as_integer_tuple(kernel_size),
     strides = as_integer_tuple(strides),
@@ -225,7 +225,7 @@ layer_conv_3d <- function(object, filters, kernel_size, strides = c(1L, 1L, 1L),
                           batch_input_shape = NULL, batch_size = NULL, dtype = NULL, 
                           name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$Conv3D, object, list(
+  create_layer(keras$layers$Conv3D, object, list(
     filters = as.integer(filters),
     kernel_size = as_integer_tuple(kernel_size),
     strides = as_integer_tuple(strides),
@@ -300,7 +300,7 @@ layer_conv_2d_transpose <- function(object, filters, kernel_size, strides = c(1L
                                     batch_input_shape = NULL, batch_size = NULL, dtype = NULL, 
                                     name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$Conv2DTranspose, object, list(
+  create_layer(keras$layers$Conv2DTranspose, object, list(
     filters = as.integer(filters),
     kernel_size = as_integer_tuple(kernel_size),
     strides = as_integer_tuple(strides),
@@ -387,7 +387,7 @@ layer_conv_3d_transpose <- function(object, filters, kernel_size, strides = c(1,
                                     kernel_constraint = NULL, bias_constraint = NULL, input_shape = NULL,
                                     batch_input_shape = NULL, batch_size = NULL, dtype = NULL, 
                                     name = NULL, trainable = NULL, weights = NULL) {
-  call_layer(keras$layers$Conv3DTranspose, object, list(
+  create_layer(keras$layers$Conv3DTranspose, object, list(
     filters = as.integer(filters),
     kernel_size = as_integer_tuple(kernel_size),
     strides = as_integer_tuple(strides),
@@ -469,7 +469,7 @@ layer_separable_conv_2d <- function(object, filters, kernel_size, strides = c(1L
                                     depthwise_constraint = NULL, pointwise_constraint = NULL, bias_constraint = NULL,
                                     batch_size = NULL, name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$SeparableConv2D, object, list(
+  create_layer(keras$layers$SeparableConv2D, object, list(
     filters = as.integer(filters),
     kernel_size = as_integer_tuple(kernel_size),
     strides = as_integer_tuple(strides),
@@ -516,7 +516,7 @@ layer_separable_conv_2d <- function(object, filters, kernel_size, strides = c(1L
 layer_upsampling_1d <- function(object, size = 2L,
                                 batch_size = NULL, name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$UpSampling1D, object, list(
+  create_layer(keras$layers$UpSampling1D, object, list(
     size = as.integer(size),
     batch_size = as_nullable_integer(batch_size),
     name = name,
@@ -552,7 +552,7 @@ layer_upsampling_1d <- function(object, size = 2L,
 layer_upsampling_2d <- function(object, size = c(2L, 2L), data_format = NULL,
                                 batch_size = NULL, name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$UpSampling2D, object, list(
+  create_layer(keras$layers$UpSampling2D, object, list(
     size = as.integer(size),
     data_format = data_format,
     batch_size = as_nullable_integer(batch_size),
@@ -598,7 +598,7 @@ layer_upsampling_2d <- function(object, size = c(2L, 2L), data_format = NULL,
 layer_upsampling_3d <- function(object, size= c(2L, 2L, 2L), data_format = NULL,
                                 batch_size = NULL, name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$UpSampling3D, object, list(
+  create_layer(keras$layers$UpSampling3D, object, list(
     size = as.integer(size),
     data_format = data_format,
     batch_size = as_nullable_integer(batch_size),
@@ -628,7 +628,7 @@ layer_upsampling_3d <- function(object, size= c(2L, 2L, 2L), data_format = NULL,
 #' @export
 layer_zero_padding_1d <- function(object, padding = 1L,
                                   batch_size = NULL, name = NULL, trainable = NULL, weights = NULL) {
-  call_layer(keras$layers$ZeroPadding1D, object, list(
+  create_layer(keras$layers$ZeroPadding1D, object, list(
     padding = as.integer(padding),
     batch_size = as_nullable_integer(batch_size),
     name = name,
@@ -667,7 +667,7 @@ layer_zero_padding_1d <- function(object, padding = 1L,
 layer_zero_padding_2d <- function(object, padding = c(1L, 1L), data_format = NULL,
                                   batch_size = NULL, name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$ZeroPadding2D, object, list(
+  create_layer(keras$layers$ZeroPadding2D, object, list(
     padding = normalize_padding(padding, 2L),
     data_format = data_format,
     batch_size = as_nullable_integer(batch_size),
@@ -714,7 +714,7 @@ layer_zero_padding_2d <- function(object, padding = c(1L, 1L), data_format = NUL
 layer_zero_padding_3d <- function(object,  padding = c(1L, 1L, 1L), data_format = NULL,
                                   batch_size = NULL, name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$ZeroPadding3D, object, list(
+  create_layer(keras$layers$ZeroPadding3D, object, list(
     padding = normalize_padding(padding, 3L),
     data_format = data_format,
     batch_size = as_nullable_integer(batch_size),
@@ -745,7 +745,7 @@ layer_zero_padding_3d <- function(object,  padding = c(1L, 1L, 1L), data_format
 #' @export
 layer_cropping_1d <- function(object, cropping = c(1L, 1L),
                               batch_size = NULL, name = NULL, trainable = NULL, weights = NULL) {
-  call_layer(keras$layers$Cropping1D, object, list(
+  create_layer(keras$layers$Cropping1D, object, list(
     cropping = as.integer(cropping),
     batch_size = as_nullable_integer(batch_size),
     name = name,
@@ -783,7 +783,7 @@ layer_cropping_1d <- function(object, cropping = c(1L, 1L),
 layer_cropping_2d <- function(object, cropping = list(c(0L, 0L), c(0L, 0L)), data_format = NULL,
                               batch_size = NULL, name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$Cropping2D, object, list(
+  create_layer(keras$layers$Cropping2D, object, list(
     cropping = normalize_cropping(cropping, 2L),
     data_format = data_format,
     batch_size = as_nullable_integer(batch_size),
@@ -834,7 +834,7 @@ layer_cropping_2d <- function(object, cropping = list(c(0L, 0L), c(0L, 0L)), dat
 #' @export
 layer_cropping_3d <- function(object, cropping = list(c(1L, 1L), c(1L, 1L), c(1L, 1L)), data_format = NULL,
                               batch_size = NULL, name = NULL, trainable = NULL, weights = NULL) {
-  call_layer(keras$layers$Cropping3D, object, list(
+  create_layer(keras$layers$Cropping3D, object, list(
     cropping = normalize_cropping(cropping, 3L),
     data_format = data_format,
     batch_size = as_nullable_integer(batch_size),
@@ -931,7 +931,7 @@ layer_conv_lstm_2d <- function(object, filters, kernel_size, strides = c(1L, 1L)
                                return_sequences = FALSE, go_backwards = FALSE, stateful = FALSE, dropout = 0.0, recurrent_dropout = 0.0,
                                batch_size = NULL, name = NULL, trainable = NULL, weights = NULL, input_shape = NULL) {
 
-  call_layer(keras$layers$ConvLSTM2D, object, list(
+  create_layer(keras$layers$ConvLSTM2D, object, list(
     filters = as.integer(filters),
     kernel_size = as_integer_tuple(kernel_size),
     strides = as_integer_tuple(strides),

diff --git a/R/layers-core.R b/R/layers-core.R
@@ -98,7 +98,7 @@ layer_dense <- function(object, units, activation = NULL, use_bias = TRUE,
                         name = NULL, trainable = NULL, weights = NULL
                         ) {
 
-  call_layer(keras$layers$Dense, object, list(
+  create_layer(keras$layers$Dense, object, list(
     units = as.integer(units),
     activation = activation,
     use_bias = use_bias,
@@ -141,7 +141,7 @@ layer_reshape <- function(object, target_shape, input_shape = NULL,
                           batch_input_shape = NULL, batch_size = NULL, dtype = NULL, 
                           name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$Reshape, object, list(
+  create_layer(keras$layers$Reshape, object, list(
     target_shape = normalize_shape(target_shape),
     input_shape = normalize_shape(input_shape),
     batch_input_shape = normalize_shape(batch_input_shape),
@@ -179,7 +179,7 @@ layer_permute <- function(object, dims, input_shape = NULL,
                           batch_input_shape = NULL, batch_size = NULL, dtype = NULL, 
                           name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$Permute, object, list(
+  create_layer(keras$layers$Permute, object, list(
     dims = as_integer_tuple(dims),
     input_shape = normalize_shape(input_shape),
     batch_input_shape = normalize_shape(batch_input_shape),
@@ -208,7 +208,7 @@ layer_permute <- function(object, dims, input_shape = NULL,
 layer_repeat_vector <- function(object, n, 
                                 batch_size = NULL, name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$RepeatVector, object, list(
+  create_layer(keras$layers$RepeatVector, object, list(
     n = as.integer(n),
     batch_size = as_nullable_integer(batch_size),
     name = name,
@@ -259,7 +259,7 @@ layer_lambda <- function(object, f, output_shape = NULL, mask = NULL, arguments
   if (identical(backend()$backend(), "theano"))
     args$output_shape = as_integer_tuple(output_shape)
 
-  call_layer(keras$layers$Lambda, object, args)
+  create_layer(keras$layers$Lambda, object, args)
 
 }
 
@@ -285,7 +285,7 @@ layer_activity_regularization <- function(object, l1 = 0.0, l2 = 0.0, input_shap
                                           dtype = NULL, name = NULL, trainable = NULL, 
                                           weights = NULL) {
 
-  call_layer(keras$layers$ActivityRegularization, object, list(
+  create_layer(keras$layers$ActivityRegularization, object, list(
     l1 = l1,
     l2 = l2,
     input_shape = normalize_shape(input_shape),
@@ -318,7 +318,7 @@ layer_masking <- function(object, mask_value = 0.0, input_shape = NULL,
                           batch_input_shape = NULL, batch_size = NULL, dtype = NULL, 
                           name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$Masking, object, list(
+  create_layer(keras$layers$Masking, object, list(
     mask_value = mask_value,
     input_shape = normalize_shape(input_shape),
     batch_input_shape = normalize_shape(batch_input_shape),
@@ -345,7 +345,7 @@ layer_masking <- function(object, mask_value = 0.0, input_shape = NULL,
 layer_flatten <- function(object, 
                           batch_size = NULL, name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$Flatten, object, list())
+  create_layer(keras$layers$Flatten, object, list())
 
 }
 
@@ -398,8 +398,21 @@ normalize_shape <- function(shape) {
 }
 
 
-# Helper function to call a layer
-call_layer <- function(layer_function, object, args) {
+#' Create a Keras Layer
+#' 
+#' @param layer_class Python layer class or R6 class of type KerasLayer
+#' @param object Object to compose layer with. This is either a 
+#' [keras_sequential_model()] to add the layer to, or another Layer which
+#' this layer will call.
+#' @param args Arguments to layer constructor function 
+#' 
+#' @return A Keras layer
+#' 
+#' @note The `object` parameter can be missing, in which case the 
+#' layer is created without a connection to an existing graph.
+#' 
+#' @export
+create_layer <- function(layer_class, object, args) {
 
   # remove kwargs that are null
   args$input_shape <- args$input_shape
@@ -410,8 +423,8 @@ call_layer <- function(layer_function, object, args) {
   args$trainable <- args$trainable
   args$weights <- args$weights
 
-  # call function
-  layer <- do.call(layer_function, args)
+  # create layer from class
+  layer <- do.call(layer_class, args)
 
   # compose if we have an x
   if (missing(object) || is.null(object))

diff --git a/R/layers-dropout.R b/R/layers-dropout.R
@@ -21,7 +21,7 @@
 layer_dropout <- function(object, rate, noise_shape = NULL, seed = NULL, 
                           batch_size = NULL, name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$Dropout, object, list(
+  create_layer(keras$layers$Dropout, object, list(
     rate = rate,
     noise_shape = normalize_shape(noise_shape),
     seed = seed,
@@ -58,7 +58,7 @@ layer_dropout <- function(object, rate, noise_shape = NULL, seed = NULL,
 layer_spatial_dropout_1d <- function(object, rate, 
                                      batch_size = NULL, name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$SpatialDropout1D, object, list(
+  create_layer(keras$layers$SpatialDropout1D, object, list(
     rate = rate,
     batch_size = as_nullable_integer(batch_size),
     name = name,
@@ -102,7 +102,7 @@ layer_spatial_dropout_1d <- function(object, rate,
 layer_spatial_dropout_2d <- function(object, rate, data_format = NULL, 
                                      batch_size = NULL, name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$SpatialDropout2D, object, list(
+  create_layer(keras$layers$SpatialDropout2D, object, list(
     rate = rate,
     data_format = data_format,
     batch_size = as_nullable_integer(batch_size),
@@ -148,7 +148,7 @@ layer_spatial_dropout_2d <- function(object, rate, data_format = NULL,
 layer_spatial_dropout_3d <- function(object, rate, data_format = NULL, 
                                      batch_size = NULL, name = NULL, trainable = NULL, weights = NULL) {
 
-  call_layer(keras$layers$SpatialDropout3D, object, list(
+  create_layer(keras$layers$SpatialDropout3D, object, list(
     rate = rate,
     data_format = data_format,
     batch_size = as_nullable_integer(batch_size),

diff --git a/R/layers-embedding.R b/R/layers-embedding.R
@@ -39,7 +39,7 @@
 layer_embedding <- function(object, input_dim, output_dim, embeddings_initializer = "uniform", embeddings_regularizer = NULL, 
                             activity_regularizer = NULL, embeddings_constraint = NULL, mask_zero = FALSE, input_length = NULL, 
                             batch_size = NULL, name = NULL, trainable = NULL, weights = NULL) {
-  call_layer(keras$layers$Embedding, object, list(
+  create_layer(keras$layers$Embedding, object, list(
     input_dim = as.integer(input_dim),
     output_dim = as.integer(output_dim),
     embeddings_initializer = embeddings_initializer,

diff --git a/R/layers-locally-connected.R b/R/layers-locally-connected.R
@@ -30,7 +30,7 @@ layer_locally_connected_1d <- function(object, filters, kernel_size, strides = 1
                                        bias_initializer = "zeros", kernel_regularizer = NULL, bias_regularizer = NULL, 
                                        activity_regularizer = NULL, kernel_constraint = NULL, bias_constraint = NULL, 
                                        batch_size = NULL, name = NULL, trainable = NULL, weights = NULL) {
-  call_layer(keras$layers$LocallyConnected1D, object, list(
+  create_layer(keras$layers$LocallyConnected1D, object, list(
     filters = as.integer(filters),
     kernel_size = as_integer_tuple(kernel_size),
     strides = as_integer_tuple(strides),
@@ -96,7 +96,7 @@ layer_locally_connected_2d <- function(object, filters, kernel_size, strides = c
                                        bias_initializer = "zeros", kernel_regularizer = NULL, bias_regularizer = NULL, 
                                        activity_regularizer = NULL, kernel_constraint = NULL, bias_constraint = NULL, 
                                        batch_size = NULL, name = NULL, trainable = NULL, weights = NULL) {
-  call_layer(keras$layers$LocallyConnected2D, object, list(
+  create_layer(keras$layers$LocallyConnected2D, object, list(
     filters = as.integer(filters),
     kernel_size = as_integer_tuple(kernel_size),
     strides = as_integer_tuple(strides),