Closed juliasilge closed 1 year ago
I may have confused two functions in the reticulate namespace when we talked: only one of which is exported. reticulate:::python_has_module()
will probe the python installation without attempting to initialize Python in the R session. Unfortunately, reticulate::py_module_available()
suffers from the same problem as tf_config()
, it will cause reticulate to initialize Python.
Perhaps the better fix would be to guard the entire example with:
skip_on_cran()
On the advice of @t-kalinowski this is a better way to check if tensorflow is installed.