Activate python environment for FastPCA

start_FastPCA_env(
  method = c("conda", "virtualenv"),
  envname = "FastPCA",
  required = TRUE
)

Arguments

method: Character string. The method to use for environment creation. Can be "conda" (default) or "virtualenv".
envname: Character string. The name of the Python environment to create/use. Defaults to "FastPCA".
required: boolean passed to reticulate

Value

Invisibly returns TRUE if setup is successful, FALSE otherwise.

Details

Using a conda environment can allow for quick testing of new versions of pytorch and tinygrad. However, it should be used with some caution. For example, if you have already used the 'rtorch' backend in some other functions, not even necessarily part of this package, using start_FastPCA_env will attach the conda environment but the modules in the environment won't be available. This is a system level conflict between something in reticulate and torch (maybe libtorch variables?). Because of this, it is really only recommended to use the python envronment if wanting to use the absolute latest version of torch (and in the future tinygrad) to compare performance (accuracy and speed). Otherwise, the rtorch implementation is likely sufficient.

Examples

if (FALSE) { # \dontrun{
  #for conda
  setup_py_env(method = "conda", envname = "FastPCA",
    python_version = "3.9", backend = "all", cuda = FALSE)
  start_fpca_env
} # }