Example of working R package PORT of Python package

[thierrymoudiki]

Been trying for years to port Python packages to R using reticulate (even using the new https://rstudio.github.io/reticulate/articles/package.html). More precisely, install like: install.packages. The success is quite random for packaging (only works in scripts and notebooks). Sometimes works on certain platforms and others won't. Looking for existing (small) R packages achieving this (or maybe I'm dumb) prowess. I guess it's 100% about environment management.

remotes::install_github("Techtonique/nnetsauce_r")

library(datasets)
# install.packages("reticulate") # Uncomment if reticulate is not installed.
library(reticulate)

# Use reticulate to access Python modules
sklearn <- import("sklearn")

X <- as.matrix(iris[, 1:4])
y <- as.integer(iris[, 5]) - 1L

n <- dim(X)[1]
p <- dim(X)[2]

set.seed(213)
train_index <- sample(x = 1:n, size = floor(0.8*n), replace = TRUE)
test_index <- -train_index

X_train <- as.matrix(iris[train_index, 1:4])
y_train <- as.integer(iris[train_index, 5]) - 1L
X_test <- as.matrix(iris[test_index, 1:4])
y_test <- as.integer(iris[test_index, 5]) - 1L

sklearn <- import("sklearn")

# ValueError: Sample weights must be 1D array or scalar
obj <- sklearn$tree$DecisionTreeClassifier()
obj2 <- nnetsauce::CustomClassifier(obj)
obj2$fit(X_train, y_train)
print(obj2$score(X_test, y_test))
print(obj2$predict_proba(X_test))