Incorrect conda environment information

[krzyslom]

Description in keras::install_keras says the following:

Keras and TensorFlow will be installed into an "r-tensorflow" virtual or conda environment.

However there is no default value set for envname parameter. This results in an installation inside the default r-reticulate environment, as mentioned on R tensorflow page. At the moment envname can be passed to the tensorflow::install_tensorflow through the ellipsis.

In the tensorflow package the default environment was changed in the 1.14.1 release. Setting envname = "r-tensorflow" helps to avoid potential conflicts if some modules are already installed inside the default environment. However if the r-reticulate env is desired as standard one then the docs should be corrected.

Please let me know if you would like me to provide a PR.

[skeydan]

You mean by updating this line:

https://github.com/rstudio/keras/blob/9d63d99c9286a9bdc6dc45716373e3b8874797a2/R/install.R#L5

Yes, please!

[t-kalinowski]

Hello,

The code that installs tensorflow and keras has been refactored.
Please update to the latest versions of reticulate, tensorflow, and keras:

install.packages("remotes")
remotes::install_github(paste0("rstudio/", c("reticulate", "tensorflow", "keras")))
reticulate::install_miniconda() # skip this if you want to self-install conda or use venv
keras::install_keras()

Please let us know if after updating you are still encountering installation issues:
https://github.com/rstudio/tensorflow/issues/new