remove unpack_x_y_sample_weight()

Author: t-kalinowski

R/dataset-utils.R

@@ -29,7 +29,8 @@
 #' @param sample_weight
 #' Sample weight for each element.
 #'
-#' @export
+#' @noRd
+# @export
 #' @family data utils
 #' @family utils
 # @seealso
@@ -75,7 +76,10 @@ function (x, y = NULL, sample_weight = NULL)
 #' @param data
 #' A list of the form `(x)`, `(x, y)`, or `(x, y, sample_weight)`.
 #'
-#' @export
+#' @noRd
+# This is removed because it has no real purpose in R. Use this instead:
+# c(x, y = NULL, sample_weight = NULL) %<-% data
+#  @export
 #' @family data utils
 #' @family utils
 #' @keywords internal

pkgdown/_pkgdown.yml

@@ -414,8 +414,8 @@ reference:
     - zip_lists
     - get_file
     - split_dataset
-    - pack_x_y_sample_weight
-    - unpack_x_y_sample_weight
+    # - pack_x_y_sample_weight
+    # - unpack_x_y_sample_weight
 
   - subtitle: "Serialization Utils"
   - contents:

tools/archive/make.R

@@ -12,12 +12,6 @@ if(!"source:tools/utils.R" %in% search()) envir::attach_source("tools/utils.R")
 ## Implemented but Questioning  ----
 
 
-# TODO: pack_x_y_sample_weight() to return an unconverted py tuple?
-#       Marked @internal for now
-#       or just remove pack_x_y_sample_weight/unpack_x_y_sample_weight from the namespace?
-#       No real utility in R since zeallot supports:
-#         c(x, y = NULL, sample_weight = NULL) %<-% data
-
 # TODO: remove usage of all.equal(<pyobj>, <pyobj>) in examples/code,
 #       export a better way.
 #       Exported all.equal S3 methods for KerasTensor and KerasVariable

vignettes-src/custom_train_step_in_tensorflow.Rmd

@@ -76,7 +76,7 @@ and we query results from `self$metrics` at the end to retrieve their current va
 CustomModel <- new_model_class(
   "CustomModel",
   train_step = function(data) {
-    c(x, y, sample_weight) %<-% unpack_x_y_sample_weight(data)
+    c(x, y = NULL, sample_weight = NULL) %<-% data
 
     with(tf$GradientTape() %as% tape, {
       y_pred <- self(x, training = TRUE)
@@ -150,7 +150,7 @@ CustomModel <- new_model_class(
     self$loss_fn <- loss_mean_squared_error()
   },
   train_step = function(data) {
-    c(x, y, sample_weight) %<-% unpack_x_y_sample_weight(data)
+    c(x, y = NULL, sample_weight = NULL) %<-% data
 
     with(tf$GradientTape() %as% tape, {
       y_pred <- self(x, training = TRUE)
@@ -212,7 +212,7 @@ it manually if you don't rely on `compile()` for losses & metrics)
 CustomModel <- new_model_class(
   "CustomModel",
   train_step = function(data) {
-    c(x, y, sample_weight) %<-% unpack_x_y_sample_weight(data)
+    c(x, y = NULL, sample_weight = NULL) %<-% data
 
     with(tf$GradientTape() %as% tape, {
       y_pred <- self(x, training = TRUE)
@@ -268,7 +268,7 @@ CustomModel <- new_model_class(
   "CustomModel",
   test_step = function(data) {
     # Unpack the data
-    c(x, y, sw) %<-% unpack_x_y_sample_weight(data)
+    c(x, y = NULL, sw = NULL) %<-% data
     # Compute predictions
     y_pred = self(x, training = FALSE)
     # Updates the metrics tracking the loss

vignettes-src/parked/_customizing_what_happens_in_fit.Rmd

@@ -216,7 +216,7 @@ it manually if you don't rely on `compile()` for losses & metrics)
 custom_model <- new_model_class(
   "custom_model",
   train_step = function(data) {
-    c(x, y, sample_weight) %<-% unpack_x_y_sample_weight(data)
+    c(x, y = NULL, sample_weight = NULL) %<-% data
 
     with(tf$GradientTape() %as% tape, {
       y_pred <- self(x, training = TRUE)
@@ -271,7 +271,7 @@ custom_model <- new_model_class(
   "custom_model",
   test_step = function(data) {
     # Unpack the data
-    c(x, y, sw) %<-% unpack_x_y_sample_weight(data)
+    c(x, y = NULL, sw = NULL) %<-% data
     # Compute predictions
     y_pred = self(x, training = FALSE)
     # Updates the metrics tracking the loss

vignettes-src/parked/_intro_to_keras_for_researchers.Rmd

@@ -826,7 +826,7 @@ CustomModel <- new_model_class(
     self$optimizer <- optimizer_adam(learning_rate=1e-3)
   },
   train_step = function(data) {
-    c(x, y, sample_weight) %<-% unpack_x_y_sample_weight(data)
+    c(x, y = NULL, sample_weight = NULL) %<-% data
     with(tf$GradientTape() %as% tape, {
       y_pred <- self(x, training=TRUE)
       loss <- self$loss_fn(y = y, y_pred = y_pred, sample_weight=sample_weight)

vignettes/custom_train_step_in_tensorflow.Rmd

@@ -78,7 +78,8 @@ and we query results from `self$metrics` at the end to retrieve their current va
 CustomModel <- new_model_class(
   "CustomModel",
   train_step = function(data) {
-    c(x, y, sample_weight) %<-% unpack_x_y_sample_weight(data)
+    # unpack data into x, y, and sample_weight
+    c(x, y = NULL, sample_weight = NULL) %<-% data
 
     with(tf$GradientTape() %as% tape, {
       y_pred <- self(x, training = TRUE)
@@ -163,7 +164,8 @@ CustomModel <- new_model_class(
     self$loss_fn <- loss_mean_squared_error()
   },
   train_step = function(data) {
-    c(x, y, sample_weight) %<-% unpack_x_y_sample_weight(data)
+    # unpack data into x, y, and sample_weight
+    c(x, y = NULL, sample_weight = NULL) %<-% data
 
     with(tf$GradientTape() %as% tape, {
       y_pred <- self(x, training = TRUE)
@@ -235,7 +237,8 @@ it manually if you don't rely on `compile()` for losses & metrics)
 CustomModel <- new_model_class(
   "CustomModel",
   train_step = function(data) {
-    c(x, y, sample_weight) %<-% unpack_x_y_sample_weight(data)
+    # unpack data into x, y, and sample_weight
+    c(x, y = NULL, sample_weight = NULL) %<-% data
 
     with(tf$GradientTape() %as% tape, {
       y_pred <- self(x, training = TRUE)
@@ -300,7 +303,7 @@ CustomModel <- new_model_class(
   "CustomModel",
   test_step = function(data) {
     # Unpack the data
-    c(x, y, sw) %<-% unpack_x_y_sample_weight(data)
+    c(x, y = NULL, sw = NULL) %<-% data
     # Compute predictions
     y_pred = self(x, training = FALSE)
     # Updates the metrics tracking the loss

vignettes/intro_to_keras_for_researchers.Rmd

@@ -1106,7 +1106,7 @@ CustomModel <- new_model_class(
     self$optimizer <- optimizer_adam(learning_rate=1e-3)
   },
   train_step = function(data) {
-    c(x, y, sample_weight) %<-% unpack_x_y_sample_weight(data)
+    c(x, y = NULL, sample_weight = NULL) %<-% data
     with(tf$GradientTape() %as% tape, {
       y_pred <- self(x, training=TRUE)
       loss <- self$loss_fn(y = y, y_pred = y_pred, sample_weight=sample_weight)