How do I install greta dependencies?

Under the hood, greta uses Google’s TensorFlow and tensorflow-probability python packages. For most users there is nothing to install: the first time you use greta in a session, it automatically installs a compatible Python, TensorFlow, and TensorFlow Probability via uv. So library(greta) followed by your first model usually just works.

If you need a conda environment instead (for example on an offline network), or want to pin specific versions, greta provides an installation helper, install_greta_deps(), which installs the needed versions into a “greta-env-tf2” conda environment. This isolates these exact python modules from other python installations, so that only greta will see them. After running it, point greta at the environment with greta_set_python("conda") and restart R. See the “Installing dependencies” vignette for the full details.

If you need to use greta without internet access, see the “Using greta offline” section of the “Installing dependencies” vignette.

How do I know if greta has the right versions of Python dependencies installed?

The most reliable check is greta_sitrep() (“situation report”). It prints the Python, TensorFlow, and TensorFlow Probability that greta has resolved, which environment they came from, and whether greta is ready to use:

library(greta)
greta_sitrep()

If everything is in place it reports the versions and tells you greta is ready; if something is missing or incompatible it tells you what.

You can also just use greta and watch what happens. Running library(greta) first prints a message about which objects are masked from base R (these only apply to greta arrays, so they don’t affect functions like %*% or rowMeans on ordinary R objects). Then run some greta code, such as:

normal(0, 1)

The first time you use greta in a session, Python is initialised and greta installs or checks its dependencies (TensorFlow and TensorFlow Probability). You’ll see a short progress message:

#> ℹ Initialising python and checking dependencies, this may take a moment.
#> ✔ Initialising python and checking dependencies ... done!

After that, your greta code returns as normal. If Python can’t be found, or the dependencies can’t be resolved, greta raises an error describing the problem (and reticulate prints the underlying resolver output above it). In that case we recommend restarting R and trying again. If the automatic setup can’t reach the internet, or you need a conda environment, run install_greta_deps() and then greta_set_python("conda"). If that does not work there is another installation approach below.

Is there an alternative way to install Python dependencies?

If the previous installation helper did not work, you can try the following:

reticulate::install_miniconda()
reticulate::conda_create(
  envname = "greta-env-tf2",
  python_version = "3.11"
)
reticulate::py_install(
  packages = c(
    "numpy",
    "tensorflow==2.15.1",
    "tensorflow-probability==0.23.0"
  ),
  envname = "greta-env-tf2",
  pip = TRUE
)

Which will install the python modules into a conda environment named “greta-env-tf2”. Point greta at it with greta_set_python("conda") and restart R.

If these instructions do not work for you, please post on the greta forum and we will respond to you as soon as we can.

I get the message: “Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX AVX2”

Briefly, this is a warning that you can safely ignore. Less briefly, it means that there can be some optimisations made with a special install of tensorflow that mean it will run faster on your machine. For more details, see this stack overflow thread. We have noted this issue in this github issue, and might in the future make it easier to resolve.