force version 2.2.2 of keras until model serailization bugs are fixed

.travis.yml

@@ -17,7 +17,6 @@ cache:
 
 before_script:
   - pip2.7 install --upgrade --ignore-installed --user travis pip setuptools wheel virtualenv
-  - pip2.7 install --upgrade --ignore-installed --user travis keras h5py pyyaml requests Pillow scipy theano
-  - R -e 'tensorflow::install_tensorflow()'
+  - pip2.7 install --upgrade --ignore-installed --user travis keras==2.2.2 h5py pyyaml requests Pillow scipy theano tensorflow
 
 

R/install.R

@@ -111,7 +111,7 @@ install_keras <- function(method = c("auto", "virtualenv", "conda"),
 
   # resolve version
   if (identical(version, "default"))
-    version <- ""
+    version <- "==2.2.2"
   else
     version <- paste0("==", version)
 

index.Rmd

@@ -46,10 +46,10 @@ This website provides documentation for the R interface to Keras. See the main K
 
 ### Installation
 
-First, install the keras R package from CRAN as follows:
+First, install the keras R package from GitHub as follows:
 
 ```{r, eval=FALSE}
-install.packages("keras")
+devtools::install_github("rstudio/keras")
 ```
 
 The Keras R interface uses the [TensorFlow](https://www.tensorflow.org/) backend engine by default. To install both the core Keras library as well as the TensorFlow backend use the `install_keras()` function: 

vignettes/getting_started.Rmd

@@ -31,10 +31,10 @@ This website provides documentation for the R interface to Keras. See the main K
 
 ## Installation
 
-First, install the keras R package from CRAN as follows:
+First, install the keras R package from GitHub as follows:
 
 ```{r, eval=FALSE}
-install.packages("keras")
+devtools::install_github("rstudio/keras")
 ```
 
 The Keras R interface uses the [TensorFlow](https://www.tensorflow.org/) backend engine by default. To install both the core Keras library as well as the TensorFlow backend use the `install_keras()` function: