Update inp_spec mapping

Author: davidrsch

R/create_keras_spec_helpers.R

@@ -125,10 +125,16 @@ collect_spec_args <- function(
 #'     to the provided string. This is the common case for blocks that take a
 #'     single tensor input.
 #' 2.  **Multiple Input Mapping**: If `input_map` is a named character vector,
-#'     it provides an explicit mapping from new argument names (the names of the
-#'     vector) to the original argument names in the `block` function (the values
-#'     of the vector). This is used for blocks with multiple inputs, like a
-#'     concatenation layer.
+#'     the **names must match the argument names of `block`** and each value
+#'     must be the name of an upstream layer block whose output should be fed
+#'     into that argument. This orientation matches the
+#'     syntax (e.g., `c(numeric = "processed_numerical")`). This is used for
+#'     blocks with multiple inputs, like a concatenation layer.
+#'
+#' _Note_: Prior releases accepted the opposite orientation
+#' (`c(processed_numerical = "numeric")`). Existing code written in that style
+#' must flip the names/values when upgrading to this version.
+
 #'
 #' @param block A function that defines a Keras layer or a set of layers. The
 #'   first arguments should be the input tensor(s).
@@ -164,7 +170,7 @@ collect_spec_args <- function(
 #'   path_b = inp_spec(dense_block, "main_input"),
 #'   concatenated = inp_spec(
 #'     concat_block,
-#'     c(path_a = "input_a", path_b = "input_b")
+#'     c(input_a = "path_a", input_b = "path_b")
 #'   ),
 #'   output = inp_spec(output_block, "concatenated")
 #' )
@@ -188,19 +194,19 @@ inp_spec <- function(block, input_map) {
     # Case 1: Single string, rename first argument
     names(new_formals)[1] <- input_map
   } else if (is.character(input_map) && !is.null(names(input_map))) {
-    # Case 2: Named vector for mapping
-    if (!all(input_map %in% original_names)) {
-      missing_args <- input_map[!input_map %in% original_names]
+    # Case 2: Named vector for mapping (argument-first)
+    if (!all(names(input_map) %in% original_names)) {
+      missing_args <- names(input_map)[!names(input_map) %in% original_names]
       stop(paste(
         "Argument(s)",
         paste(shQuote(missing_args), collapse = ", "),
         "not found in the block function."
       ))
     }
-    # Use match() for a more concise, vectorized replacement of names
+
     new_names <- original_names
-    match_indices <- match(input_map, original_names)
-    new_names[match_indices] <- names(input_map)
+    match_indices <- match(names(input_map), original_names)
+    new_names[match_indices] <- unname(input_map)
     names(new_formals) <- new_names
   } else {
     stop("`input_map` must be a single string or a named character vector.")

README.md

@@ -1,13 +1,14 @@
-
 # kerasnip
 
 <!-- badges: start -->
+
 [![Lifecycle: experimental](https://img.shields.io/badge/lifecycle-experimental-orange.svg)](https://lifecycle.r-lib.org/articles/stages.html#experimental)
 [![R-CMD-check](https://github.com/davidrsch/kerasnip/actions/workflows/R-CMD-check.yaml/badge.svg)](https://github.com/davidrsch/kerasnip/actions/workflows/R-CMD-check.yaml)
 [![Codecov test
 coverage](https://codecov.io/gh/davidrsch/kerasnip/graph/badge.svg)](https://app.codecov.io/gh/davidrsch/kerasnip)
 [![CRAN_Status_Badge](https://www.r-pkg.org/badges/version/kerasnip)](https://cran.r-project.org/package=kerasnip)
 [![Downloads](https://cranlogs.r-pkg.org/badges/kerasnip)](https://cran.r-project.org/package=kerasnip)
+
 <!-- badges: end -->
 
 The goal of `kerasnip` is to provide a seamless bridge between the `keras` and `tidymodels` frameworks. It allows for the dynamic creation of `parsnip` model specifications for Keras models, making them fully compatible with `tidymodels` workflows.
@@ -112,8 +113,7 @@ create_keras_functional_spec(
     main_input = input_block,
     path_a = inp_spec(path_block, "main_input"),
     path_b = inp_spec(path_block, "main_input"),
-    concatenated = inp_spec(concat_block, c(path_a = "input_a", path_b = "input_b")),
-    # The output block must be named 'output'.
+    concatenated = inp_spec(concat_block, c(input_a = "path_a", input_b = "path_b")),
     output = inp_spec(output_block, "concatenated")
   ),
   mode = "regression"
@@ -136,6 +136,7 @@ fit(spec, mpg ~ ., data = mtcars) |>
 #> 4  18.6
 #> 5  17.9
 ```
+
 ### Example 3: Tuning a Sequential MLP Architecture
 
 This example demonstrates how to tune the number of dense layers and the rate of a final dropout layer, showcasing how to tune both architecture and block hyperparameters simultaneously.
@@ -210,11 +211,11 @@ tune_res <- tune_grid(
 # 6. Show the best architecture
 show_best(tune_res, metric = "rmse")
 #> # A tibble: 5 × 7
-#>   num_dense dense_units dropout_rate .metric .estimator .mean .config              
-#>       <int>       <int>        <dbl> <chr>   <chr>      <dbl> <chr>                
+#>   num_dense dense_units dropout_rate .metric .estimator .mean .config
+#>       <int>       <int>        <dbl> <chr>   <chr>      <dbl> <chr>
 #> 1         1          64          0.1 rmse    standard    2.92 Preprocessor1_Model02
 #> 2         1          64          0.5 rmse    standard    3.02 Preprocessor1_Model08
 #> 3         3          64          0.1 rmse    standard    3.15 Preprocessor1_Model04
 #> 4         1           8          0.1 rmse    standard    3.20 Preprocessor1_Model01
 #> 5         3           8          0.1 rmse    standard    3.22 Preprocessor1_Model03
-```
+```

man/inp_spec.Rd

@@ -28,7 +28,7 @@ test_that("E2E: Functional spec (classification) works", {
       path_b = inp_spec(path_block, "main_input"),
       concatenated = inp_spec(
         concat_block,
-        c(path_a = "input_a", path_b = "input_b")
+        c(input_a = "path_a", input_b = "path_b")
       ),
       output = inp_spec(output_block_class, "concatenated")
     ),
@@ -165,7 +165,7 @@ test_that("E2E: Multi-input, single-output functional classification works", {
       path_b = inp_spec(dense_path, "flatten_b"),
       concatenated = inp_spec(
         concat_block,
-        c(path_a = "in_1", path_b = "in_2")
+        c(in_1 = "path_a", in_2 = "path_b")
       ),
       output = inp_spec(output_block_class, "concatenated")
     ),
@@ -230,7 +230,7 @@ test_that("E2E: Functional spec with pre-constructed optimizer works", {
       path_b = inp_spec(path_block, "main_input"),
       concatenated = inp_spec(
         concat_block,
-        c(path_a = "input_a", path_b = "input_b")
+        c(input_a = "path_a", input_b = "path_b")
       ),
       output = inp_spec(output_block_class, "concatenated")
     ),
@@ -283,7 +283,7 @@ test_that("E2E: Functional spec with string loss works", {
       path_b = inp_spec(path_block, "main_input"),
       concatenated = inp_spec(
         concat_block,
-        c(path_a = "input_a", path_b = "input_b")
+        c(input_a = "path_a", input_b = "path_b")
       ),
       output = inp_spec(output_block_class, "concatenated")
     ),

tests/testthat/test_e2e_func_classification.R

@@ -23,7 +23,7 @@ test_that("E2E: Functional spec (regression) works", {
       path_b = inp_spec(path_block, "main_input"),
       concatenated = inp_spec(
         concat_block,
-        c(path_a = "input_a", path_b = "input_b")
+        c(input_a = "path_a", input_b = "path_b")
       ),
       output = inp_spec(output_block_reg, "concatenated")
     ),
@@ -74,7 +74,7 @@ test_that("E2E: Functional regression works with named predictors in formula", {
       path_b = inp_spec(path_block, "main_input"),
       concatenated = inp_spec(
         concat_block,
-        c(path_a = "input_a", path_b = "input_b")
+        c(input_a = "path_a", input_b = "path_b")
       ),
       output = inp_spec(output_block_reg, "concatenated")
     ),
@@ -185,7 +185,7 @@ test_that("E2E: Multi-input, multi-output functional regression works", {
       path_b = inp_spec(dense_path, "input_b"),
       concatenated = inp_spec(
         concat_block,
-        c(path_a = "in_1", path_b = "in_2")
+        c(in_1 = "path_a", in_2 = "path_b")
       ),
       output_1 = inp_spec(output_block_1, "concatenated"),
       output_2 = inp_spec(output_block_2, "concatenated")

tests/testthat/test_e2e_func_regression.R

@@ -16,6 +16,35 @@ test_that("inp_spec throws error for mismatched input_map names", {
 })
 
 
+test_that("inp_spec supports argument-first mapping", {
+  block_with_args <- function(numeric, categorical) {
+    list(numeric = numeric, categorical = categorical)
+  }
+  mapper <- c(
+    numeric = "processed_numeric",
+    categorical = "processed_categorical"
+  )
+  wrapped <- kerasnip:::inp_spec(block_with_args, mapper)
+
+  expect_identical(
+    names(formals(wrapped))[1:2],
+    c("processed_numeric", "processed_categorical")
+  )
+  res <- wrapped(processed_numeric = 10, processed_categorical = 20)
+  expect_identical(res$numeric, 10)
+  expect_identical(res$categorical, 20)
+})
+
+test_that("inp_spec rejects the legacy input_map orientation", {
+  block_with_args <- function(input_a, input_b) {}
+  legacy_mapper <- c(processed_a = "input_a", processed_b = "input_b")
+  expect_error(
+    kerasnip:::inp_spec(block_with_args, legacy_mapper),
+    "not found in the block function"
+  )
+})
+
+
 test_that("inp_spec throws error for invalid input_map type", {
   block_with_args <- function(a) {}
   expect_error(

tests/testthat/test_inp_spec.R

@@ -115,10 +115,10 @@ create_keras_functional_spec(
     processed_2 = inp_spec(dense_path_2, "input_2"),
     concatenated = inp_spec(
       concat_block,
-      c(processed_1 = "input_a", processed_2 = "input_b")
+      c(input_a = "processed_1", input_b = "processed_2")
     ),
     output_1 = inp_spec(output_block_1, "concatenated"), # New output block 1
-    output_2 = inp_spec(output_block_2, "concatenated")  # New output block 2
+    output_2 = inp_spec(output_block_2, "concatenated") # New output block 2
   ),
   mode = "regression" # Still regression, but will have two columns in y
 )
@@ -159,7 +159,7 @@ train_df <- tibble::tibble(
     function(i) x_data_2[i, , drop = FALSE]
   ),
   output_1 = y_data_1, # Named output 1
-  output_2 = y_data_2  # Named output 2
+  output_2 = y_data_2 # Named output 2
 )
 
 rec <- recipe(output_1 + output_2 ~ input_1 + input_2, data = train_df)

vignettes/functional_api.Rmd

@@ -32,10 +32,10 @@ First, we load the necessary packages.
 library(kerasnip)
 library(tidymodels)
 library(keras3)
-library(dplyr)       # For data manipulation
-library(ggplot2)     # For plotting
-library(future)      # For parallel processing
-library(finetune)    # For racing
+library(dplyr) # For data manipulation
+library(ggplot2) # For plotting
+library(future) # For parallel processing
+library(finetune) # For racing
 ```
 
 ## Data Preparation
@@ -149,10 +149,10 @@ create_keras_functional_spec(
     combined_features = inp_spec(
       concatenate_features,
       c(
-        processed_numerical = "numeric",
-        processed_neighborhood = "neighborhood",
-        processed_bldg = "bldg",
-        processed_condition = "condition"
+        numeric = "processed_numerical",
+        neighborhood = "processed_neighborhood",
+        bldg = "processed_bldg",
+        condition = "processed_condition"
       )
     ),
     output = inp_spec(output_regression, "combined_features")
@@ -288,7 +288,7 @@ final_ames_fit |>
   plot(show_shapes = TRUE)
 ```
 
-![Model](images/model_plot_shapes_fs.png){fig-alt="A picture showing the model shape"}
+![Model](images/model_plot_shapes_wf.png){fig-alt="A picture showing the model shape"}
 
 ```{r inspect-final-keras-model-history}
 # Plot the training history

vignettes/workflows_functional.Rmd

@@ -32,10 +32,10 @@ First, we load the necessary packages.
 library(kerasnip)
 library(tidymodels)
 library(keras3)
-library(dplyr)          # For data manipulation
-library(ggplot2)        # For plotting
-library(future)         # For parallel processing
-library(finetune)       # For racing
+library(dplyr) # For data manipulation
+library(ggplot2) # For plotting
+library(future) # For parallel processing
+library(finetune) # For racing
 ```
 
 ## Data Preparation