search

StatCan / aaw-kubeflow-containers

Containers built to be used with Kubeflow for Data Science
Other
24 stars 21 forks source link

Investigate upgrading ‘tidymodels’ / Upgrade R #485

Closed bryanpaget closed 1 year ago

bryanpaget commented 1 year ago

I am able to install the latest version of tidymodels==1.1.0 but I am unable to import the package.

> library("tidymodels")
Error: package or namespace load failed for ‘tidymodels’ in loadNamespace(i, c(lib.loc, .libPaths()), versionCheck = vI[[i]]):
 namespace ‘rlang’ 1.0.6 is being loaded, but >= 1.1.0 is required

In order to upgrade this package we may need to upgrade R.

bryanpaget commented 1 year ago

I've noticed we are installing R in the base image but then also using R from Anaconda.

bryanpaget commented 1 year ago

We updated the base of the docker image to use the latest R (4.3.1 as of this writing) but our Anaconda environment is stuck on 4.1.3:

image

bryanpaget commented 1 year ago

I will investigate updating the Anaconda base environment to 4.3.1.

bryanpaget commented 1 year ago

We only install Anaconda on the remote desktop image which means all other images have an unmanaged Anaconda installation. I moved the Anaconda installation block from the remote desktop Dockerfile and I moved it into one of the dockerbits shared by all images. This way we can update the Anaconda environment for all images in one go resulting in greater consistency and we also get more control over versioning.

bryanpaget commented 1 year ago

This issue has uncovered another issue: we don't have a consistent Anaconda installation across our images and all images seem to be stuck using an older version of Anaconda. Perhaps the Anaconda version is being pinned somewhere. The work of updating Anaconda in order to have the latest R packages such as tidymodels, will be tracked with the following ticket: https://github.com/StatCan/aaw-kubeflow-containers/issues/496