Development on this package is not active. Please take a look at https://github.com/mlr-org/mlr3keras for an implementation of tabnet on mlr3 via keras instead.
The original intention for this package was to use fast.ai tabular and a pytorch implementation of TabNet, in R as an mlr3 learner.
However torch crashes Rstudio at time of writing due to rstudio/reticulate#273.
Consequently we look to develop the requisite functionality in keras which is better supported in R instead.