Error loading saved ranger model after updating mlr3verse package

[Schwarzeneggerjune]

Description:
I encountered an issue after updating the mlr3verse package in R. I created a binary classification model using ranger with mlr3verse version 0.2.8 in R 4.3.0 and saved the model as an .rds file for future use. However, after updating one of the packages within mlr3verse, I can no longer load the saved model file. The error message is as follows:
LearnerClassifRanger:classif.ranger

• Model: ranger
  Error in private$.get_values() : attempt to apply non-function
  I attempted to resolve the issue by reinstalling the previous version of mlr3verse (0.2.8), but the problem persists. I suspect that the update might have caused a compatibility issue with the saved model.
  Steps to Reproduce:
  Create a binary classification model using ranger with mlr3verse version 0.2.8.
  Save the model as an .rds file using saveRDS().
  Update one of the packages within mlr3verse.
  Load the saved model using readRDS() and attempt to use it.
  Expected behavior:
  The model should load correctly and be usable as before the update.
  Actual behavior:
  The model fails to load with the error message mentioned above.
  System Information:
  R version: 4.3.0
  mlr3verse version: 0.2.8 (initially) and updated versions
  Platform: [Your operating system, e.g., Windows 10, macOS Monterey]
  Additional context:
  I have tried the following steps to resolve the issue:
  Reinstalling the previous version of mlr3verse (0.2.8).
  Checking for any changes in the ranger package dependencies.
  However, none of these steps have resolved the issue. I would appreciate any guidance on how to resolve this compatibility problem or if there is a way to convert the saved model to a compatible format.
  Thank you for your assistance!

[Schwarzeneggerjune]

the model was stored at https://github.com/Schwarzeneggerjune/ColonTrack

[be-marc]

Can you try https://mlr-org.com/faq.html#old-model-broken?

[Schwarzeneggerjune]

Can you try https://mlr-org.com/faq.html#old-model-broken?

Yes，actually， I have tried this method. My model was built based on mlr3 0.16.1, I tried to re-install mlr3 0.16.1, but it didn't work. And then I built a new model with the lastest mlr3 and run the following code.
learner = lrn("classif.rpart")
learner_old = readRDS("learner.rds")
learner$state = learner_old$state
I used the new learner to predict, there is still error,
learner$predict(task$new)
error:Setting row roles 'test'/'holdout' is no longer possible.

[Schwarzeneggerjune]

Can you try https://mlr-org.com/faq.html#old-model-broken?

I have fixed this issue. The error was generated from the R6 class ParaSet, and this class was defined in package 'paradox'. So actually the error is not due to the incorrect version of mlr3, but rather due to the incorrect version of paradox. After installing the old version of paradox, the model ran successfully.