Textrecipes series: TF-IDF | Emil Hvitfeldt

[Isaiah]

Hi,

I've been working through Emil's recent blog post: https://www.hvitfeldt.me/blog/textrecipes-series-tfidf/

If I comment out

doParallel::registerDoParallel()

things run fine. If I leave

doParallel::registerDoParallel() 

in the script, the line

lasso_grid <- tune_grid(
  lasso_wf,
  resamples = wine_boot,
  grid = param_grid
)

fails with

Warning message:
All models failed in tune_grid(). See the `.notes` column.

and

lasso_grid$.notes
[[1]]
# A tibble: 1 x 1
  .notes                                                              
  <chr>                                                               
1 "recipe: Error in UseMethod(\"prep\"): no applicable method for 'pr~

The following diagnostics work:

doParallel::registerDoParallel()
getDoParName()

"doParallelSNOW"

getDoParWorkers()

3

getDoParVersion()

"1.0.15"

detectCores()
8

And

library(doParallel)
cl <- makeCluster(2)
registerDoParallel(cl)
foreach(i=1:3) %dopar% sqrt(i)

runs fine

I'm using R3.6, doParallel 1.0.15, tidymodels 0.1.0, recipes 0.1.12, tune 0.1.0 and RStudio 1.3.959
OS is Windows 10.

[Emilhvitfeldt]

It appears that this is an issue with how windows handles parallel computing in R.

A similar issue popped up on stackoverflow which might be helpful.

[Isaiah]

Thanks Emil!

The error in the StackOverflow question looks the same.

I tried

all_cores <- parallel::detectCores(logical = FALSE)
cl <- makePSOCKcluster(all_cores)
registerDoParallel(cl)

and

library(doFuture)
registerDoFuture()
all_cores <- parallel::detectCores(logical = FALSE)
cl <- makeCluster(all_cores)
plan(cluster, workers = cl)

though I get the same error in each case.