Bump required R version to 3.4

Author: t-kalinowski

.github/workflows/R-CMD-check.yaml

@@ -31,7 +31,7 @@ jobs:
           - {os: 'ubuntu-20.04', tf: 'default', r: '3.6'} # default R in ubuntu-20.04
           - {os: 'ubuntu-20.04', tf: 'default', r: '3.5'} #
           - {os: 'ubuntu-20.04', tf: 'default', r: '3.4'} #
-          - {os: 'ubuntu-20.04', tf: 'default', r: '3.3'} # Oldest R available in Rstudio Marketplace offerings
+          # - {os: 'ubuntu-20.04', tf: 'default', r: '3.3'} # Oldest R available in Rstudio Marketplace offerings; rcmdcheck not available
           # - {os: 'ubuntu-20.04', tf: 'tensorflow-cpu', r: '3.2'} # rcmdcheck not available
           # - {os: 'ubuntu-20.04', tf: 'tensorflow-cpu', r: '3.1'} # remotes not available
 

DESCRIPTION

@@ -26,7 +26,7 @@ License: MIT + file LICENSE
 URL: https://keras.rstudio.com
 BugReports: https://github.com/rstudio/keras/issues
 Depends:
-    R (>= 3.2)
+    R (>= 3.4)
 Imports:
     generics (>= 0.0.1),
     reticulate (>= 1.10),

NEWS.md

@@ -53,9 +53,9 @@
 
 - Fixed an issue in `layer_input()` where passing a tensorflow `DType` objects to argument `dtype` would throw an error.
 
-- Fixed an issue in `compile()` where passing an R function via an in-line 
+- Fixed an issue in `compile()` where passing an R function via an in-line
   call would result in an error from subsequent `fit()` calls.
-  (e.g., `compile(loss = function(y_true, y_pred) my_loss(y_true, y_pred))` 
+  (e.g., `compile(loss = function(y_true, y_pred) my_loss(y_true, y_pred))`
   now succeeds)
 
 - `clone_model()` gains a `clone_function` argument that allows you to customize each layer as it is cloned.
@@ -75,6 +75,7 @@
 - `%py_class%` gains the ability to delay initializing the python session until first use.
   It is now safe to implement and export `%py_class%` objects in an R package.
 
+- Bumped minimum R version to 3.4. Expanded CI to test on all supported R version. Fixed regresison that prevent package installation on R <= 3.4
 # keras 2.6.0
 
 Breaking changes (Tensorflow 2.6):