don't install tensorflow-metal by default

Author: t-kalinowski

NEWS.md

@@ -3,11 +3,11 @@
 - Updates for reticulate 1.41. The tensorflow R package now calls
   `reticuate::py_require()` when it is loaded. Calling `install_tensorflow()`
   in most circumstances is no longer necessary.
-- GPU usage on M-series Macs is once again enabled by default.
 - `install_tensorflow()` installs TensorFlow v2.18 by default.
 - Fixed an issue where GPUs would not be found when running on Windows
   WSL Linux (reported in rstudio/keras3#1456, fixed in #599)
 - Fixes for NumPy 2.0 (#601)
+- Fixes for R-devel (4.5)
 
 # tensorflow 2.16.0
 

R/install.R

@@ -1,6 +1,7 @@
 #' Install TensorFlow and its dependencies
 #'
 #' @description
+#'
 #' Beginning with reticulate version 1.41, in most circumstances, calling the
 #' `install_tensorflow()` function is no longer necessary, because reticulate
 #' automatically registers python requirements with `reticulate::py_require()`
@@ -12,13 +13,16 @@
 #' - On Linux: if a GPU is detected: `"tensorflow[and-cuda]"`, otherwise,
 #' `"tensorflow-cpu"`.
 #'
-#' - On macOS: `c("tensorflow", "tensorflow-metal")`. To prevent TensorFlow usage of
-#' the GPU, call `reticulate::py_require("tensorflow-metal", action = "remove")`
-#' before reticulate has initialized Python.
+#' - On macOS: `"tensorflow"` is declared. The default package is not capable
+#' of using the GPU. To enable TensorFlow usage of the GPU, call
+#' `reticulate::py_require("tensorflow-metal")` before reticulate has
+#' initialized Python. Note that not all features of TensorFlow work correctly
+#' if `tensorflow-metal` is installed. There are known issues with random number
+#' generators like `tf$random$stateless_uniform()`, likely others as well.
 #'
 #' - On Windows: `"tensorflow"` is declared. Note that TensorFlow GPU usage on
-#' Windows is no longer supported. To use a GPU on windows, use TensorFlow via
-#' WSL.
+#' Windows is no longer supported (Since TensorFlow 2.10). To use a GPU on
+#' windows, use TensorFlow via WSL.
 #'
 #' `install_tensorflow()` creates a new virtual environment containing the
 #' `tensorflow` python package and it's direct dependencies. For creating a

man/install_tensorflow.Rd

@@ -24,6 +24,7 @@ test_that("tf.random works", {
     dtype = tf$dtypes$int32)
   expect_s3_class(x, "tensorflow.tensor")
   x <- as.array(x)
-  expect_type(x, "integer")
+  if (!is_windows())
+    expect_type(x, "integer")
   expect_identical(dim(x), 10L)
 })