renaming extract_keras_summary to extract_keras_model

Author: davidrsch

R/keras_tools.R

@@ -33,17 +33,24 @@ keras_evaluate <- function(object, x, y = NULL, ...) {
   keras3::evaluate(keras_model, x = x_proc, y = y_proc, ...)
 }
 
-#' Extract Keras Model Summary
+#' Extract the Raw Keras Model from a Kerasnip Fit
 #'
+#' @title Extract Keras Model from a Fitted Kerasnip Object
 #' @description
-#' Extracts and returns the summary of a Keras model fitted with `kerasnip`.
+#' Extracts and returns the underlying Keras model object from a `parsnip`
+#' `model_fit` object created by `kerasnip`.
+#'
+#' @details
+#' This is useful when you need to work directly with the Keras model object for
+#' tasks like inspecting layer weights, creating custom plots, or passing it to
+#' other Keras-specific functions.
 #'
 #' @param object A `model_fit` object produced by a `kerasnip` specification.
-#' @param ... Additional arguments passed on to `keras3::summary()`.
 #'
-#' @return A character vector, where each element is a line of the model summary.
+#' @return The raw Keras model object (`keras_model`).
+#' @seealso keras_evaluate, extract_keras_history
 #' @export
-extract_keras_summary <- function(object, ...) {
+extract_keras_model <- function(object) {
   object$fit$fit
 }
 

tests/testthat/test_e2e_features.R

@@ -139,7 +139,7 @@ test_that("E2E: Setting num_blocks = 0 works for sequential models", {
   fit_obj <- parsnip::fit(spec, mpg ~ ., data = mtcars)
 
   # Check that the dense layer is NOT in the model
-  keras_model <- fit_obj |> extract_keras_summary()
+  keras_model <- fit_obj |> extract_keras_model()
   expect_equal(length(keras_model$layers), 1) # Output layers only
 
   # Check layer names explicitly
@@ -162,7 +162,7 @@ test_that("E2E: Error handling for reserved names works", {
   )
 })
 
-test_that("E2E: extract_keras_summary works", {
+test_that("E2E: extract_keras_model works", {
   skip_if_no_keras()
 
   # Reuse model setup from previous tests
@@ -194,7 +194,7 @@ test_that("E2E: extract_keras_summary works", {
 
   fit_obj <- parsnip::fit(spec, mpg ~ ., data = mtcars)
 
-  summary_output <- extract_keras_summary(fit_obj)
+  summary_output <- extract_keras_model(fit_obj)
 
   expect_type(summary_output, "closure")
   expect_true(any(grepl("Layer ", summary_output)))

tests/testthat/test_e2e_functional.R

@@ -210,7 +210,7 @@ test_that("E2E: Block repetition works for functional models", {
     set_engine("keras")
   fit_1 <- fit(spec_1, mpg ~ ., data = mtcars)
   model_1_layers <- fit_1 |>
-    extract_keras_summary() |>
+    extract_keras_model() |>
     pluck("layers")
 
   # Expect 3 layers: Input, Dense, Output
@@ -221,7 +221,7 @@ test_that("E2E: Block repetition works for functional models", {
     set_engine("keras")
   fit_2 <- fit(spec_2, mpg ~ ., data = mtcars)
   model_2_layers <- fit_2 |>
-    extract_keras_summary() |>
+    extract_keras_model() |>
     pluck("layers")
   # Expect 4 layers: Input, Dense, Dense, Output
   expect_equal(length(model_2_layers), 4)
@@ -231,7 +231,7 @@ test_that("E2E: Block repetition works for functional models", {
     set_engine("keras")
   fit_3 <- fit(spec_3, mpg ~ ., data = mtcars)
   model_3_layers <- fit_3 |>
-    extract_keras_summary() |>
+    extract_keras_model() |>
     pluck("layers")
   # Expect 2 layers: Input, Output
   expect_equal(length(model_3_layers), 2)

vignettes/getting_started.Rmd

@@ -181,12 +181,13 @@ mlp_fit <- fit(mlp_spec, y ~ x, data = train_df)
 
 ```{r model-summarize}
 mlp_fit |> 
-  extract_keras_summary()
+  extract_keras_model() |> 
+  summary()
 ```
 
 ```{r model-plot, eval=FALSE}
 mlp_fit |> 
-  extract_keras_summary() |> 
+  extract_keras_model() |> 
   plot(show_shapes = TRUE)
 ```
 
@@ -320,7 +321,8 @@ We can now inspect our final, tuned model.
 # Print the model summary
 final_fit |>
   extract_fit_parsnip() |>
-  extract_keras_summary()
+  extract_keras_model() |> 
+  summary()
 
 # Plot the training history
 final_fit |>